Node Catalog

Explore 86 node references across knowledge, automation, motion, simulation, 3D, and media. Open a node's full reference for its connections, parameter options, buttons, and usage steps. Search by name, control, or capability, or choose a category.

Browse every node below, or choose one from the node list.

Core and Utility

For Each

for_each

Loop over Mocap Collection outputs, checked BVH paths, or a numeric parameter range. The start node passes the active animation downstream.

Use and connections: Connect Collection output to source, route the animation through processing and Scene to Render, and connect Render to for_each_end.render. Use Run For Each, Stop, Loop Limit, Single pass, or Start at iteration. A comment wrapper groups the loop; each render must finish before the next iteration.

Workflow guide →
Inputs, parameters, buttons, and usage

Connections and result

Connect a Mocap Collection to source. Send the start node output through animation processing, Scene, and Render. Connect Render to the paired For Each End render input.

Parameters and options

ControlWhat it does
IterateCollection outputs respects the collection’s Checked only setting. Checked paths always uses checked files. Parameter range changes a numeric parameter instead of selecting takes.
Loop Limit0 / All visits every eligible input. A positive number caps the iterations run, including when starting partway through the list.
Single pass / iterationRuns one eligible iteration; numbering starts at 1. This overrides Loop Limit and counts eligible inputs, not the original collection row numbers.
Loop Start / start iterationStarts at a chosen eligible iteration and continues. Enabling it turns off Single pass.
Parameter / Count / First value / StepIn Parameter range mode, choose the parameter to vary, number of values, initial value, and increment.

Buttons and editing actions

ControlWhat it does
Run For EachStarts the loop and waits for each connected render to complete before advancing.
StopStops the running loop.
Create wrapperRecreates the comment wrapper and paired end node if the wrapper was deleted.

Using this node

  1. Load and check takes in Mocap Collection.
  2. Connect the render chain and its completion output to For Each End.
  3. Try Single pass first, then run the full range.

Putting a Render node inside the comment wrapper does not replace the required connection to the end node.

For Each End

for_each_end

Complete a For Each iteration when its connected Render job finishes.

Use and connections: Use the paired end node created with For Each. Connect the Render node to its render input; grouping nodes visually is not enough to define render completion.

Workflow guide →
Inputs, parameters, buttons, and usage

Connections and result

Connect the final Render node to render. This is the completion boundary for the paired For Each start node.

Parameters and options

ControlWhat it does
renderThe render job that must finish before the start node advances.
Iteration statusReports loop progress; this is runtime state rather than a second loop counter to configure.

Buttons and editing actions

ControlWhat it does
Paired loop controlsUse Run For Each and Stop on the start node. The end node supplies the completion connection.

Using this node

  1. Create a For Each node and use its paired end node.
  2. Connect Render output to render on the end node.
  3. Run a single iteration to confirm the render finishes before advancing.

note

Note tasks and text content with per-parameter controls.

Inputs, parameters, buttons, and usage

Connections and result

Use a Note to supply text or task parameters to downstream nodes. Gantt Chart can turn connected Note parameters into scheduled tasks.

Output and checks: Note keeps all regular nonempty values available, with their parameter names as metadata, whether checked or unchecked. Prompt selection happens in the receiving Qwen Image or MiniMax H3 node: a Note with multiple visible parameters contributes only checked values, without parameter labels. A Note with exactly one visible parameter contributes its value even unchecked. Hidden/internal parameters do not count; an empty second visible parameter still makes it a multi-parameter Note. H3 also honors these selections through Append and Switch; an empty connected selection reports an error instead of using a saved local prompt.

Parameters and options

ControlWhat it does
Parameter nameIdentifies a text value or task. Use distinct names when matching tasks between charts.
Parameter valueContains the text sent downstream. Select a parameter in the info panel to edit its contents.
Featured parameterChooses which parameter is expanded in the node body; other parameters remain on the node.
Row checkboxMarks a value for consumers that use selection, including Qwen Image and all MiniMax H3 prompt inputs. It does not remove unchecked data from the Note's general output.

Buttons and editing actions

ControlWhat it does
Add Param / Remove SelectedAdds a parameter row or removes the selected row in the info panel.
Apply ChangesSaves the edited parameter names and values to the graph.
Edit Params… / Rename…Opens parameter editing or changes the node name.
Set display to selected paramSelect a row in the parameter table, then use this info-panel button to show its value inside the Note.

Using this node

  1. Add named parameters and enter their text.
  2. Select the most useful row and set it as the displayed parameter.
  3. Connect the Note to Append, Librarian, or a Gantt Chart.

node

Standard general-purpose parameter node.

Inputs, parameters, buttons, and usage

Connections and result

A general-purpose graph node carries named parameters and links. Choose a specialized node when you need a renderer, model request, or file-processing action.

Parameters and options

ControlWhat it does
Name and parameter namesGive the node and each value meaningful names so downstream connections can be identified.
Parameter valuesEdit values in the parameter/info panel. Their meaning depends on the node receiving them.
ConnectionsRoute the desired value to a downstream input; a generic node has no specialized conversion settings.

Buttons and editing actions

ControlWhat it does
Add ParamAdds a named value in the info-panel parameter table.
Remove SelectedRemoves the selected parameter row.
Apply ChangesSaves the edited names and values to the graph.
Edit Params… / Rename…Opens parameter editing or changes the node name. There is no node-specific generation button.

Using this node

  1. Create a Node and add the values needed by your workflow.
  2. Name the values clearly.
  3. Connect them to the receiving node and configure that node for the operation.

append

Ordered input aggregation; can drive Gantt group order.

Inputs, parameters, buttons, and usage

Connections and result

Connect several upstream text sources. Append combines their content in its displayed input order; Gantt Chart also uses this order for groups.

Parameters and options

ControlWhat it does
Input orderOrder matters: the preview and saved result follow the Append list, not alphabetical node names.
Upstream parameter textEach entry identifies the source node/parameter and its resolved text. Empty text contributes no content.
Local and connected valuesThe merge preview distinguishes the entries collected for this Append node.

Buttons and editing actions

ControlWhat it does
Preview Merge (Append order)Opens the collected entries and merged text for inspection.
Copy TextCopies the merged text from the preview dialog.
Save JSONWrites the collected merge data to a chosen JSON file.

Using this node

  1. Connect the notes or text-producing nodes to combine.
  2. Arrange their order in Append.
  3. Preview the merge before connecting it to a prompt or Gantt Chart.

switch

Single-branch selection from multiple upstream nodes.

Inputs, parameters, buttons, and usage

Connections and result

Connect alternative branches and route the selected branch to a downstream node.

Parameters and options

ControlWhat it does
Active branchOnly the selected input branch supplies the active value downstream.
Input orderThe branch selector follows the connected branch list. Check the displayed source when stepping through alternatives.

Buttons and editing actions

ControlWhat it does
Previous branch / Next branchThe left and right arrow buttons change the selected input.

Using this node

  1. Connect two or more alternative sources.
  2. Use the arrows to select the desired source.
  3. Check the downstream preview before running its operation.

gantt_chart

Scheduling board with multi-note grouping, collapse state persistence, and CSV copy/paste transfer.

Inputs, parameters, buttons, and usage

Connections and result

Connect Notes directly or through Append. Each connected Note becomes a task group; Append determines group order.

Parameters and options

ControlWhat it does
Task rows and datesAssign work to days on the chart. The horizontal day scrollbar moves the visible date range.
ProgressTrack completion for individual tasks alongside their scheduled days.
GroupsExpand or collapse each Note group. The group state is saved with the workflow.
Reminder date / time / messageSelect a task, choose a notification date and 12-hour time with AM/PM, and enter the reminder message.
Task panel widthResize the task-name area to leave enough room for labels.

Buttons and editing actions

ControlWhat it does
TodayJumps the chart back to the current date.
CopyCopies assignments and progress and writes a CSV transfer log.
PasteMatches task names and transfers their assignments and progress to this chart; unrelated tasks remain.
Reminder controlOpens reminder scheduling for the selected task.

Using this node

  1. Connect Notes containing named tasks.
  2. Assign dates and progress in the chart.
  3. Copy and paste between charts when their task names should share a schedule.

Copy/paste matches task names. Use consistent names for tasks that should receive the same schedule.

python

Run custom scripts in graph context.

Inputs, parameters, buttons, and usage

Connections and result

Connected text is available to the script through inputs and primary_input. The script also receives node, params, raw_params, and graph_scene.

Parameters and options

ControlWhat it does
CodeThe Python source executed by this node. Edit and save it before running.
params / raw_paramsRead named parameter values or the original parameter entries.
inputs / primary_inputinputs includes source-node keys and in1, in2-style keys. primary_input is the first nonempty resolved text input.
output_text / resultSet output_text for the downstream text; result is used next, then captured standard output when neither is set.

Buttons and editing actions

ControlWhat it does
Edit Code…Opens the node’s Python script editor.
Run PythonExecutes the script and displays captured output or errors in the info-panel console.
Add to ShelfSaves this node as a user-invoked shelf tool through the app.

Using this node

  1. Connect source text and open Edit Code.
  2. Read primary_input and assign the transformed text to output_text.
  3. Run Python and inspect the console and downstream value.

output

Terminal collector for final graph output.

Inputs, parameters, buttons, and usage

Connections and result

Connect the final branch to this terminal node. Running it refreshes the active upstream path and its displayed result.

Parameters and options

ControlWhat it does
Mode: AutoUpdates automatically as the relevant graph state changes.
Mode: ManualLeaves execution under the Run button’s control.
Mode: Run On StartupRuns the output path on startup of the loaded workflow.

Buttons and editing actions

ControlWhat it does
RunRecomputes the active path and refreshes the output display and connected response widgets.

Using this node

  1. Connect the final processing node to Output.
  2. Choose Manual while assembling the graph.
  3. Press Run, then select the automatic mode appropriate to the workflow.

AI and Knowledge

local_server

Local server endpoint configuration.

Inputs, parameters, buttons, and usage

Connections and result

Embeds the configured server webpage in the node. It is a browser-facing endpoint node; LLM Prompt provides provider/model request settings.

Parameters and options

ControlWhat it does
url / address / endpointSet the server address using one of these recognized parameter names. Changing it updates the embedded browser target.
Embedded pageControls inside the browser belong to the server you opened and depend on that service.

Buttons and editing actions

ControlWhat it does
Page interactionUse the embedded webpage’s controls. This node does not install or start the external server for you.

Using this node

  1. Start the local service you want to use.
  2. Enter its address on the node.
  3. Confirm the embedded page loads and use its own interface.

The embedded page requires Qt WebEngine. An unavailable service or incorrect URL will not become reachable simply by adding this node.

llm_prompt

OpenAI or Ollama/DeepSeek model settings and response generation for Chatbot workflows.

Inputs, parameters, buttons, and usage

Connections and result

Connect prompt text and optional files; connect this node to Chatbot to supply its model configuration. output_path can specify where standalone output is saved.

Parameters and options

ControlWhat it does
ProviderChoose OpenAI, DeepSeek, or Ollama. Provider selection changes the relevant connection and model controls.
ModelSelect or enter a model supported by the chosen provider. The configured default is gpt-4.1-mini.
TemperatureControls generation variability; the node default is 0.2. Supported behavior depends on the selected model.
API keyCredentials for the selected hosted provider.
Ollama URLAddress of the local Ollama server; default http://localhost:11434. The model must be installed on that server.
prompt / files / output_pathProvide the request, optional referenced files, and optional saved-output location. Connected inputs supply the values used by the request.

Buttons and editing actions

ControlWhat it does
Send to GPTCollects the prompt and file inputs and sends them to the selected provider; the button also serves Ollama and DeepSeek.

Using this node

  1. Choose the provider and a model available there.
  2. Provide credentials or a working Ollama address.
  3. Connect a Note to prompt and send a small request before adding files.

chatbot

Conversation workflows with Database history, Voice Actor input, LLM Prompt runtime, and optional mediator_input.

Inputs, parameters, buttons, and usage

Connections and result

Connect llm_prompt for generation settings, database for stored chat history, Voice Actor to the voice input, and optional mediator_input for agent tool handoff.

Parameters and options

ControlWhat it does
Message entryText for the next conversation turn.
Model connectionThe linked LLM Prompt node supplies provider, model, credentials, and generation settings.
Database connectionThe connected Database node chooses the stored conversation context.
Voice and Mediator connectionsVoice Actor can supply recognized speech. Mediator receives supported tool requests when connected.

Buttons and editing actions

ControlWhat it does
SendSubmits the entered message using the configured conversation and model context.
RefreshReloads the displayed conversation state.

Using this node

  1. Connect and configure LLM Prompt.
  2. Optionally select a Database collection/project.
  3. Enter a message and Send; add voice or agent handoff after the text path works.

Chatbot owns its spoken response when Voice Actor is connected to it; adding Mediator is not a second independent speaking path.

database

Persistence layer for graph and chat data.

Inputs, parameters, buttons, and usage

Connections and result

The mongo_uri, collection, project, and note inputs configure a MongoDB-backed context that can be connected to Chatbot.

Parameters and options

ControlWhat it does
Mongo URIMongoDB connection string; default mongodb://localhost:27017. The database service must be reachable.
Collections / Collection nameChoose an existing collection or enter the name to use.
Projects / Project nameChoose or name the project within the collection.
Note textText to store in the selected collection/project.

Buttons and editing actions

ControlWhat it does
RefreshReloads available database selections.
Create/UseCreates or selects the entered context.
Save NoteStores the entered note in the chosen context.
Delete Project / Delete CollectionRemoves the selected stored project or collection. Check the selection before using these actions.

Using this node

  1. Enter a reachable Mongo URI and Refresh.
  2. Choose a collection and project, or enter names and Create/Use.
  3. Save a note or connect the configured Database to Chatbot.

data_nexus

Folder-aware knowledge canvas for points, vault files, graph relationships, question and slide point types, review rings, references, and answer-to-question links.

Inputs, parameters, buttons, and usage

Connections and result

Outputs a structured knowledge graph for Skills, Task, and Mediator. The canvas organizes points, folders, links, questions, and generated slides.

Parameters and options

ControlWhat it does
Point and folder selectionSelect the item whose content or relationships you want to edit. Folder organization also controls the vault representation.
Needs? / Refine?Show or hide source-point lines for questions needing stronger answers or refinement.
Pull / maximum link stretchPull linked points together only after a link exceeds the chosen stretch distance.
ZoomUse −, 1:1, and + to inspect the graph without changing its data.
Point typesQuestion, refinement, slide, and ordinary point types have distinct colors. White review rings identify newly added points.

Buttons and editing actions

ControlWhat it does
Add point / Create folderAdds knowledge or an organizational folder.
Link selected pointCreates a relationship from the selected point.
Delete selected pointRequests deletion of the selected point and asks first.
Arrange / ClusterRepositions linked points or separates point types for review.
Save / Open-Switch vault / Merge vaultSaves the sidecar or selects and combines vault data.
Reload / Open folderReloads vault files or opens their folder for inspection.

Using this node

  1. Create folders and factual points.
  2. Link related points and arrange the canvas.
  3. Connect Data Nexus to Task and review generated questions or slides before generating another draft.

skills

Local Skills library browser for human templates, agent templates, task guides, approval state, validation warnings, review diffs, and Sales Agent deck generation.

Inputs, parameters, buttons, and usage

Connections and result

Connect Mediator for template/guide creation and Data Nexus for generation. The skills output supplies the selected library asset to Task.

Parameters and options

ControlWhat it does
Library selectionChoose the human template, agent template, task guide, or agent guide to work with.
Version and approval stateLatest, active approved, draft, and deprecated entries describe different versions and readiness states.
Validation resultsCheck required metadata and referenced resources before using a template for generation.

Buttons and editing actions

ControlWhat it does
RefreshRescans the local Skills library.
Create Template / Create GuideUses Teacher Agent to derive an agent template from a human template, or a task guide from an agent template.
Generate DeckUses an approved Sales Agent template and connected Data Nexus to create an HTML deck.
Review DiffCompares the selected version with the active approved or latest version.
Approve / Draft / DeprecateChanges the selected asset’s lifecycle state.

Using this node

  1. Refresh and select a template.
  2. Review validation and the version diff before approving it.
  3. Connect the selection and Data Nexus to Task, or generate a deck directly from the approved template.

task

Workflow step controller that reads Skills, Data Nexus, and optional Mediator AI; checks readiness, writes questions, generates drafts, refreshes HTML, and indexes slides.

Inputs, parameters, buttons, and usage

Connections and result

Connect skills, data_nexus, and optionally mediator. The node exposes status, questions, and artifact outputs for the task workflow.

Parameters and options

ControlWhat it does
Task statusSet Draft, Approved, Dismissed, or another supported workflow state; this is separate from merely generating a draft.
Template and task guideThe selected Skills assets define the instructions and expected output.
Questions and reportInspect missing facts, refinement questions, and readiness feedback before the next step.
ArtifactThe latest generated output can feed an HTML Preview path input.

Buttons and editing actions

ControlWhat it does
Run StepChecks connected assets, facts, and Mediator readiness.
Write QuestionsWrites preparation/refinement questions into Data Nexus without generating a deck.
Generate DraftAsks the Sales Agent to create or update deck copy and indexes the resulting slides.
Refresh HTMLRebuilds the current deck using the latest style without rewriting its copy.
Index SlidesReads the existing deck and creates or updates its Data Nexus slide points.

Using this node

  1. Choose an approved Skills template and add supporting facts to Data Nexus.
  2. Run Step and answer outstanding questions.
  3. Generate Draft, review it in HTML Preview, and then set the appropriate task status.

librarian

Document indexing, search, summarize, and analysis.

Inputs, parameters, buttons, and usage

Connections and result

The query, docs_dir, mode, top_k, and action inputs configure retrieval or analysis. Connected text takes priority over the corresponding local value.

Parameters and options

ControlWhat it does
queryQuestion or search text. Note and Append can provide longer queries.
docs_dirDirectory containing the documents to work with.
modeChoose tree_summarize, the default, or compact for the Librarian response-synthesis mode.
top_kLimits the number of retrieved results used by the operation; default 5.
actionSupports search, analyze, analyze_with_sources, summarize, and summarize_with_sources.

Buttons and editing actions

ControlWhat it does
SearchRetrieves material matching the query.
AnalyzeRequests analysis of the query and available document context.
Analyze (with Sources)Requests an analysis that includes source references.

Using this node

  1. Configure the document folder and Librarian service.
  2. Connect query text or enter it locally.
  3. Search first to check coverage, then analyze with sources when you need traceable support.

Automation, Voice, and Control

voice_actor

Speech-to-text and text-to-speech node with Auto Respond, Ask First, and Manual send modes.

Inputs, parameters, buttons, and usage

Connections and result

text supplies speech content; audio can receive Audio Capture. The text output carries recognized speech to Chatbot, Translator, or Mediator.

Parameters and options

ControlWhat it does
Voice → Text / Text → VoiceSwitch between recognition and speaking. Some connected automatic response sources force Text → Voice.
Parameter selectorChooses the connected parameter to speak when several text values are available.
Voice / M-F / LangChoose an installed or configured voice, applicable gender selection, and recognition/voice language. Availability depends on the installed speech backend.
STTSelect the recognition engine; local Whisper and Google have different dependencies.
Send modeAuto Respond sends after speech pauses. Ask First confirms before sending. Manual sends when Stop is pressed. Connected live capture can stream queued transcript chunks.

Buttons and editing actions

ControlWhat it does
Listen / SpeakStarts the operation for the current mode; the same main control can replay the latest automatic response.
Pause / SendPauses listening or sends the transcript, depending on the active send mode.
Replay / UndoReplays spoken output or undoes the last transcript chunk in the applicable mode.
Stop / ClearStops the active operation or clears the transcript when idle. In Manual recognition mode, stopping sends the transcript.

Using this node

  1. Choose the mode, language, and speech backend.
  2. For captured system audio, connect Audio Capture to audio.
  3. Listen and inspect the transcript, or connect text and test Speak.

audio_capture

Capture output-device, input-device, or application audio with source, backend, sample-rate, threshold, silence, and phrase controls. Continuous streaming supports buffered speech recognition, exclusion of Qubit Field playback, and Test & Listen monitoring.

Inputs, parameters, buttons, and usage

Connections and result

The audio output supplies captured audio to Voice Actor or another compatible audio consumer. This node chooses the source; the consumer starts ongoing capture.

Parameters and options

ControlWhat it does
Source / device / applicationChoose output-device audio, an input device such as a microphone, or application audio, then select the relevant device/process.
BackendSelect the capture backend; auto lets the node choose.
Sample rateCapture sampling rate; default 16000 Hz.
ThresholdAudio level used to identify speech/activity; default 0.012.
Silence / phrase durationSilence ends a phrase; maximum phrase length bounds a chunk. Defaults are 850 ms silence and 20000 ms maximum phrase duration.
LiveKeeps capture waiting for speech and breaks long speech into shorter live chunks.
Exclude QubitMCP audio (all outputs)Captures other applications while excluding this app’s speech and test playback. It ignores the output-device selector and needs supported Windows process-loopback capture. Restart listening after changing it.

Buttons and editing actions

ControlWhat it does
RefreshReloads available capture sources.
Test & ListenRecords three seconds, reports its level, then plays the sample after recording has stopped.

Using this node

  1. Select a source and run Test & Listen.
  2. Adjust the level threshold if the sample is too quiet or noisy.
  3. Connect audio to Voice Actor and start listening there.

The level meter is in dBFS: 0 is full scale and more negative values are quieter.

translator

Translate incoming text or voice transcripts with language, style, and automatic translation controls.

Connect: Send text to text and connect an llm_prompt node to the matching input to choose the provider and model. The translated result leaves through text and can feed a Voice Actor for speech.

Use: Select From and To languages, choose Natural or Literal style, and enable Auto for new incoming text or click Translate manually.

Inputs, parameters, buttons, and usage

Connections and result

Connect text to text and LLM Prompt to llm_prompt. The text output carries the translation, for example into Voice Actor.

Parameters and options

ControlWhat it does
From / ToSource and target languages; automatic source detection and Mandarin Chinese are the initial settings.
StyleNatural favors fluent wording; Literal stays closer to the source construction.
AutoTranslates new incoming text automatically when enabled.
Model settingsProvider, credentials, and model come from the connected LLM Prompt node.

Buttons and editing actions

ControlWhat it does
TranslateRuns a translation manually with the current text and language settings.
Debug controlWrites Translator diagnostics for troubleshooting.

Using this node

  1. Configure LLM Prompt and connect it.
  2. Choose From, To, and Style.
  3. Enter or connect text and Translate before enabling Auto.

openai_realtime_translate

Translate live microphone speech with spoken output and a live translation transcript.

Setup: Enter an OpenAI API key, select the model and voice, and set the source and target languages. This node captures microphone audio directly and requires the PyAudio and websockets dependencies.

Use: Click Listen to begin. Configure bidirectional translation or add a second language route, and adjust the silence interval for speech turns. The node exposes audio and text outputs.

Inputs, parameters, buttons, and usage

Connections and result

Captures the microphone directly and exposes audio and text outputs. It requires the node’s realtime connection and microphone dependencies.

Parameters and options

ControlWhat it does
API key / modelConfigure access to the realtime model used by this node.
Source / target languageChoose the language route; initial settings are English to Mandarin Chinese.
BidirectionalAllows the configured language pair to translate in both directions.
Second routeAn additional source/target language route can be added.
VoiceSelect marin or cedar in the node’s voice control.
SilenceSets the speech-turn silence interval; initial value is 500 ms.

Buttons and editing actions

ControlWhat it does
+Adds another language route.
Listen / StopStarts or ends live microphone translation.

Using this node

  1. Configure credentials, model, and a language route.
  2. Set the silence interval to suit speech pauses.
  3. Press Listen and monitor the transcript and spoken translation.

minimax_tts_api

Generate speech audio from text through the MiniMax Speech API.

Connect: Feed the text input or enter text in the node. Connected text takes priority. Supply a MiniMax API key and choose a speech model and voice ID, including a custom voice ID when available.

Controls: Set speed, volume, pitch, output format, sample rate, and output folder. The generated file is available through audio and audio_path for playback or downstream audio effects.

Inputs, parameters, buttons, and usage

Connections and result

Connect text or type it in the node. A connected input overrides typed text. audio and audio_path expose the generated result.

Parameters and options

ControlWhat it does
Model / voice IDChoose the speech model and voice. Custom or cloned voice IDs can be entered.
API key / base / endpoint / timeoutConfigure the speech service connection. The node’s endpoint default is /v1/t2a_v2 and timeout is 120 seconds.
Speed / volume / pitchAdjust delivery speed, loudness, and pitch shift. Initial speed and volume are 1; pitch shift is 0.
Format / sample rate / bitrate / channelsChoose output encoding. Supported format choices include MP3, WAV, and FLAC.
Output folder / historyChoose where generated files are saved and select an existing result to replay.

Buttons and editing actions

ControlWhat it does
GenerateRequests speech for the current text.
Play / StopPlays or stops the selected audio file.
RefreshRefreshes generated-file history.
OpenOpens the output location.

Using this node

  1. Enter credentials and select a model/voice.
  2. Generate a short text sample and play it.
  3. Connect the audio path downstream once its format and sound are suitable.

mediator_agent

Agent profile layer and technical executor for direct voice workflows, Chatbot-origin tool handoff, and Sales Agent AI task review.

Inputs, parameters, buttons, and usage

Connections and result

Inputs include voice_input, chatbot_history, system_prompt, data_nexus, and sandbox. The selected agent profile interprets requests and coordinates the supported tools.

Parameters and options

ControlWhat it does
Agent profile / system promptChoose the role and instructions used to interpret incoming work.
Command entryA manual command for the agent instead of waiting for connected input.
AutoProcesses fresh incoming work automatically; disabling it lets you inspect inputs first.
Sandbox connectionSupplies the configuration for supported coding-tool execution.
Output/log viewShows progress, responses, and failures so the requested action can be reviewed.

Buttons and editing actions

ControlWhat it does
Run CmdProcesses the manually entered command.
Process InputsProcesses the current connected input context.
StopStops the current operation.
Clear / Open FolderClears the displayed log or opens the agent working/output folder.
Approval dialogsGrant/Deny and deletion approval controls are presented for actions requiring a decision.

Using this node

  1. Select the agent role and connect its required context.
  2. Process a simple request manually and inspect the result.
  3. Enable Auto when incoming work should be handled as it arrives.

codex_sandbox

Configure the local coding runner used by Mediator, including account, model, sandbox, approval, working directory, network, and output capture settings.

Inputs, parameters, buttons, and usage

Connections and result

Connect this configuration node to Mediator’s sandbox input. It supplies the local coding runner settings; it is not a standalone prompt/chat node.

Parameters and options

ControlWhat it does
Runner / executable / Codex Home / account profileChoose how the local CLI is launched and which account state it uses.
Model / reasoning effortSelect the configured coding model and effort for the runner.
Sandbox mode / approval policyControls the runner’s filesystem mode and when it requests approvals.
Project root / working directory / writable rootsSet the project context, execution directory, and additional permitted write locations.
Network access / output captureChoose network and captured-output behavior supported by the runner.
Advanced / extra argumentsExposes CLI paths, roots, and additional runner arguments.

Buttons and editing actions

ControlWhat it does
Detect CLISearches common installation locations and PATH for the executable.
StatusChecks login status using the selected CLI and home.
Login / LogoutSigns in through a terminal or logs out of the selected account state.
Browse controlsChoose the relevant executable and directory paths.

Using this node

  1. Detect the CLI and check Status.
  2. Set project paths and the runner policy you intend to use.
  3. Connect it to Mediator before running a coding request.

qubit_deck_controller

List or execute Qubit Deck actions from manual fields or Mediator JSON.

Inputs, parameters, buttons, and usage

Connections and result

Use manual fields or connect Mediator JSON to mediator_input. The controller sends requests to the configured Qubit Deck API.

Parameters and options

ControlWhat it does
API baseAddress of the running Qubit Deck API.
Button / button name / slotIdentify the button to resolve or invoke; inspect the returned label/thumbnail to confirm it.
ActionSelect the supported request, such as invoke, highlight, context listing, or health check.
ModeChoose the controller/executor behavior offered by the node; executor accepts structured requests.

Buttons and editing actions

ControlWhat it does
Run ActionSends the selected action with the current target fields.
InvokeInvokes the resolved deck button.
Open DebuggerOpens the deck debugging interface.

Using this node

  1. Point API base at the running deck service.
  2. Resolve the target by name or slot and inspect its preview.
  3. Invoke the button or connect Mediator requests for automated control.

keyboard_sequence

Keyboard Auto Runner for timed key/action sequences with lead-in, hold time, delays, and injection mode.

Inputs, parameters, buttons, and usage

Connections and result

Runs a sequence of keyboard, mouse, text, delay, and loop actions. A connected Serial COM/Pico path can provide supported hardware injection.

Parameters and options

ControlWhat it does
Action type / key / textChoose what each step does. Key/text lists can supply a different value on each loop pass.
Lead-in / key hold / delay afterSets the time before execution, how long keys remain pressed, and pauses between actions.
Delay tagsNamed delays can be reused for action delays and hold times.
Mouse target / X / Y / button / click timeConfigure a captured click location and whether coordinates are tied to a window.
Loop ID / run count / conditionsControls repeated blocks and actions restricted to particular loop passes.
Injection modeSelect the supported key-injection method for the target application.

Buttons and editing actions

ControlWhat it does
Add / Insert / Remove ActionEdits the action sequence; insertion uses the selected position.
Make LoopCreates loop start/end blocks around a selection or at the end.
Add Delay TagEdits shared timing values.
Capture ClickCaptures the target used by a mouse action.
Run Sequence / StopStarts or stops execution.
Load JSON / Save JSON / Copy output logLoads or saves the action list and copies the execution log.

Using this node

  1. Build a short sequence and set a lead-in long enough to focus the target.
  2. Test the sequence, checking key holds and mouse coordinate mode.
  3. Add loops and shared delay tags after individual actions work.

serial_com

Stores COM port, baud, and timeout settings for serial/Pico HID automation paths.

Inputs, parameters, buttons, and usage

Connections and result

Stores the serial connection used by supported keyboard/Pico automation. Select the correct device before downstream hardware actions.

Parameters and options

ControlWhat it does
PortThe COM device to communicate with.
BaudSerial communication speed; it must match the device firmware.
TimeoutHow long to wait for a response before the request is treated as timed out.

Buttons and editing actions

ControlWhat it does
RefreshRescans available COM ports.
PingChecks communication with the selected device.
Install pyserial / Copy CommandWhen the dependency is absent, install it through the app or copy the provided command.

Using this node

  1. Connect the device and Refresh.
  2. Select its COM port and matching baud/timeout.
  3. Ping before using it with Keyboard Sequence.

3D, Rigging, and Scene Processing

Look At

look_at

Aim or constrain a scene object to another object or a skeleton joint.

Use and connections: Connect an object, choose the target owner, and configure aim/up axes, look-at or copy-rotation mode, position matching, offsets, and live updating. Use joint targets for camera or object tracking in character scenes.

Workflow guide →
Inputs, parameters, buttons, and usage

Connections and result

Connect the object to constrain to object and send its object output into Scene. In Scene, assign an object or skeleton joint as the target owner.

Parameters and options

ControlWhat it does
Target ownerDrag an object or joint into the Scene target row, or use a joint path such as Character/Skeleton/Head.
Aim axis / Up axisChoose the object’s forward and up axes; defaults are −Z and Y.
Rotation mode / Match rotationUse look-at aiming or copy-rotation behavior and choose whether rotation is applied.
Match position / Position axesEnable positional following and choose the affected X/Y/Z axes.
Maintain offset / Position offsetPreserve the existing positional relationship or apply the stored offset.
Live updateWhen enabled in Scene, follows viewport target movement immediately; otherwise updates on timeline frames and Scene rebuilds.

Buttons and editing actions

ControlWhat it does
Scene target and live controlsAssign the target and toggle live updating in the Scene outliner. This node does not have a separate generation step.

Using this node

  1. Connect a camera or other object through Look At to Scene.
  2. Assign the target object or joint.
  3. Check the aim/up axes and enable only the rotation/position behavior you need.

Animation Blend

animation_blend

Build a transition from a character bind pose into a skeletal animation, useful for settling a garment before motion begins.

Use and connections: Connect the character to bind_pose and a compatible animation to animation. Set blend frames, hold frames, FPS, and curve; enable the target-frame option to start from a later pose. Send the animation output to Scene or the cloth collider path. Retarget incompatible rigs first.

Workflow guide →
Inputs, parameters, buttons, and usage

Connections and result

Connect a character reference pose to bind_pose and a compatible animation to animation. Send the animation output to Scene or an animated cloth collider path.

Parameters and options

ControlWhat it does
Blend framesTransition length from the starting pose to the animation; default 30 frames. Zero skips the transition.
Hold pose framesKeeps the starting pose before blending, useful for letting cloth settle; default 0.
Use target frame / Target frameOverrides the animation’s starting pose with a later frame; frame 0 is the first frame.
Frame rateUsed for durations and target-frame interpretation; default 24 FPS. Match the Scene timeline.
Blend curveChoose Smooth for an eased transition or Linear for a constant-rate blend; the default is Smooth.

Buttons and editing actions

ControlWhat it does
ViewPreviews the resulting animation in the viewport.

Using this node

  1. Retarget the animation first if it uses a different skeleton.
  2. Connect the reference pose and animation.
  3. Set a hold and blend duration, then View and scrub the transition.

Delete Joint

delete_joint

Remove selected joints from a derived FBX or BVH skeleton and preview the result.

Use and connections: Connect a supported rig to rig, choose Select joints, and inspect with View before using the result downstream. The derived rig leaves the imported source file intact.

Workflow guide →
Inputs, parameters, buttons, and usage

Connections and result

Connect the source rig to rig. The node produces a derived skeleton with selected joints removed, leaving the imported file unchanged.

Parameters and options

ControlWhat it does
Joint selectionSelect joints in the skeleton list or preview before deleting/restoring them.
Removed jointsThe saved selection records the joints excluded from the derived rig.
PreviewInspect the resulting skeleton before connecting it to later rig processing.

Buttons and editing actions

ControlWhat it does
Select joints…Opens the joint-selection editor.
Delete selected / Restore selectedUpdates the exclusion set for the selected joints.
Restore allClears all joint exclusions.
ViewShows the modified skeleton in the main viewport.

Using this node

  1. Connect an FBX/BVH rig and open Select joints.
  2. Remove only the joints you intend to exclude.
  3. View the result and verify downstream mapping before exporting.

Mocap Collection

mocap_collection

Organize BVH takes with an active output, Use checkboxes, Checked only filtering, and live preview. Load individual files or scan a folder recursively.

Use and connections: Use Update paths after moving source files. Load renders and Review renders associate visual reviews with takes; Rename and Archive manage BVH run folders and assigned render folders. Output a selected take to retargeting, or connect the collection to For Each for a render batch.

Workflow guide →
Inputs, parameters, buttons, and usage

Connections and result

The output supplies the active BVH take. Connect the collection to retargeting for one take or For Each for a render batch.

Parameters and options

ControlWhat it does
Use checkboxesMark takes eligible for checked-file processing; row selection alone does not change these checks.
Output indexSelects the active take. With Checked only enabled, it indexes the eligible list while the displayed path number identifies the original row.
Checked onlyRestricts the output slider to checked takes, in list order.
Live viewUpdates the viewport as the active output changes while retaining the current camera.
Render thumbnailsAssigned renders let you review a take visually before choosing it.

Buttons and editing actions

ControlWhat it does
Load BVH files / Load BVH folderAdds individual files or scans a folder recursively; existing paths are skipped.
Remove / Update pathsRemoves the active collection path or relocates selected files beneath a new parent folder.
Check all / Uncheck all / Select allChanges eligibility checks or row selection; these are separate operations.
Load renders / Review rendersAssociates generated renders and opens their review.
Rename / ArchiveRenames the active BVH run folder and assigned render folders, or moves selected runs and renders into an archive.
View / Open in ExplorerPreviews the active animation or opens its file location.

Using this node

  1. Load a BVH folder and preview several takes.
  2. Check the takes you want to process and enable Checked only.
  3. Connect the collection to For Each and run a test render.

Rename and Archive operate on folders/files, not just the displayed list labels.

Animation Correction

animation_correction

Manually correct an animated BVH or FBX take with joint position and rotation keys and body or limb IK ranges. Save per-take drafts and export an approved corrected BVH.

Manual correction workflow →
Inputs, parameters, buttons, and usage

Connections and result

Connect an animated BVH or FBX source, Mocap Collection, or Rig Limiter to animation. The animation output supplies the clip with saved corrections to downstream nodes, including Scene and Anim Retarget. Scene displays the corrected skeleton; use Anim Retarget to apply it to a character mesh.

Controls and editing actions

ControlWhat it does
Preview / edit animationOpens the input take in the 3D viewport. Select a joint in the viewport or embedded hierarchy, then scrub and pose it. The panel's ? button opens editing hints.
IK Controls / FK keyingEnables pelvis, chest, wrist, and ankle controls. Dragging previews a pose without creating keys. With IK off, select a joint and use Set key to key its local position or rotation; the selected timeline channels determine which axes are keyed. Local FK edits can also override an IK pose.
Start IK / End IK / Set keyStart and End bound an overwrite, marked S and E above the triangle keys and yellow between them. Set key adds intermediate IK poses inside a range. Select Px, Py, or Pz before Start to limit the target coordinates; separate position axes can overlap in time and share boundary frames. Clear the channel selection for full IK keying, including chain rotations.
Join IK ranges (Fuse icon)Between Start IK and End IK. Select an End and the next Start on the same control with matching channels and limb-lock settings; the right range must have its End. Joining fills the gap and keeps both selected keys as intermediate poses. The button is disabled and desaturated without a valid selection. Also available by right-click or Ctrl+J.
Show collidersDisplays body and hand capsule guides when the input rig supports them. A Rig Limiter input supplies its saved limits and volumes.
Reload all / selected joint / selected axesWith no joint selected, restores the entire input take, including IK and regular edits. With a joint selected, restores that joint and removes IK ranges driving it. Select timeline channels to restore only those axes across the take. The reset is saved to the draft immediately and can be undone in the current session.
Save draftSaves corrections beside the source as .bvh.animation_correction.json or .fbx.animation_correction.json. Keyframe actions and Reload also save the draft. Saved corrections are tied to the source take and base clip; changing the base animation can invalidate them.
Approve & export BVHSaves the draft and writes a uniquely named _corrected_ BVH beside the source. Keeps the original animation and records the approved export.

Using this node

  1. Select a collection take and, if using Rig Limiter, finish its automatic repair first.
  2. Open Preview / edit animation, scrub to the affected frames, and adjust joints or key an IK range.
  3. Save the draft, review the motion, then approve and export or pass the saved corrected animation to Scene or Anim Retarget.

Use Settings > Joint names to display joint labels. Undo and redo restore correction edits; a viewport drag is one undo step. Drafts remain specific to each source take; save the current edits before changing previews. Unkeyed IK previews are not included in drafts or BVH exports. Use the timeline curve editor to fit or frame keys, pan and scale the graph, edit tied or independent handles, and expand or detach the panel.

Rig Limiter

rig_limiter

Repair existing PriorMDM/HumanML animation with joint limits and optional body/finger collision correction. Preview a derived clip or export new BVHs without replacing the source files.

Repair workflow →
Inputs, parameters, buttons, and usage

Connections and result

Connect a compatible Mocap Import or Mocap Collection to animation. The animation output supplies a corrected skeleton clip to Anim Retarget, Animation Blend, Transforms, Animation Correction, or Scene. It preserves root motion, timing, and bone lengths; it does not output a skinned character mesh.

Use the original PriorMDM/HumanML joint names, hierarchy, and rest axes, with fixed bone lengths and no bone scaling. Apply this node before retargeting. Supported finger joints are processed when present.

Parameters and options

ControlWhat it does
Correct capsule intersectionsEnables contact correction between the tapered body, palm, and finger volumes when IK repair is also enabled. Enabled initially.
IK repairAttempts constrained finger adjustments before arm correction, and supports limb contact repair. Disable it to apply only bend/twist limits. Enabled initially.
Thorough repair (slower)Spends more time resolving between-frame contacts. Off by default for fast repairs. Both modes report unresolved contacts.
Bend / Twist or SpreadSelect joints in the preview to edit the supported angular ranges, including finger bending and base spread.
Capsule radius: Start end / Far endSizes each end independently in input rig units. Pick joints or capsules to inspect the matching controls. Foot capsules run from ankle to foot; select the ankle to adjust them in either Preview Rig or Rig Limiter.
Collision preview (blocks rotations)Shows collision volumes and stops manual rotations at contact. Save the preview panel to store its limits, radii, and collision option on this node.

Buttons and editing actions

ControlWhat it does
PreviewOpens the current source or available repaired clip without running a repair. Inspect joints and adjust limits and capsule volumes in the preview panel.
RepairTurns red when the input or saved settings need repair. For a connected BVH import or collection, checks for matching saved repairs in the background as the path or settings change, without requiring Preview. Click to calculate with progress/cancellation controls and report corrected frames, finger capsules, arm IK adjustments, and remaining overlap.
Export corrected BVHRequires an up-to-date Repair result, then writes it to a chosen folder with a unique _limited_ filename. Preserves hierarchy, root motion, timing, and End Sites.
Export checked collectionRequires a directly connected Mocap Collection. Repairs its checked takes into new files while preserving collection entries and source BVHs. Completed files remain if the batch is interrupted.
Save in the preview panelStores this node's joint limits, capsule radii, and collision option without repairing the animation. Save the graph to persist them. The previous repair stays visible until you click Repair again; compatible saved BVH repairs are restored when their source and settings still match. Otherwise the source is shown until repaired.
Reset limits and volumesRestores the default limits and capsule sizes. Run Repair again to apply them.

Repair progress includes solving and saving the corrected BVH. Stop, Escape, or closing the progress dialog requests cancellation. Matching saved repairs can be recovered after reopening the workflow without recalculating; compatible repairs saved by earlier versions are also recognized when their source and settings match.

Using this node

  1. Connect an original PriorMDM/HumanML clip or collection before Anim Retarget.
  2. Click Repair, then Preview to inspect body and finger contact, and adjust limits or capsule sizes as needed.
  3. Review remaining overlaps, then connect the animation downstream or export the active/checked takes.

Capsules are collision approximations. Unreachable contacts remain reported instead of relaxing joint limits, stretching bones, or moving the root. Inspect the retargeted character mesh before cloth simulation or rendering.

PriorMDM

priormdm

Generate a humanoid motion take from text with duration, seed, guidance, and device controls. Preview the reconstructed BVH or add completed takes to a Mocap Collection.

Use and connections: Run Setup PriorMDM and configure its runtime/model paths. Preview Rig exposes bend/twist limits and collision guides. Optional Animate Hands uses a separate Setup HandMDM runtime, side, description, time window, and blend controls. Finger and palm retargeting remain preview features.

Workflow guide →
Inputs, parameters, buttons, and usage

Connections and result

Generates a BVH motion take from text. Completed results can become MocapBVH nodes or entries in Mocap Collection.

Parameters and options

ControlWhat it does
Prompt / DurationDescribe the movement and requested clip length; initial duration is 5 seconds.
Seed / GuidanceSeed controls repeatability for the same setup; guidance changes how strongly generation follows the text. Initial values are 42 and 2.5.
DeviceChoose cuda:0 or cpu according to the configured runtime.
Preview proportionsChanges the exported preview proportions while preserving original motion data. Reference proportions can require foot/contact cleanup.
Collision volumes / Joint limitsChoose which guides Preview Rig displays for inspection.
Add hand articulation / Hand / Hand descriptionEnables the separate HandMDM pass and chooses side and gesture text.
Gesture start / end / BlendLimits the hand gesture’s time window and transition.
Advanced runtime pathsSet body/hand repositories, inference Python executables, checkpoints, normalization folder, download cache, and output folder.

Buttons and editing actions

ControlWhat it does
Setup PriorMDM / Setup HandMDMManages the separate body and finger-motion runtimes/checkpoints.
Generate / StopStarts generation or stops the current job.
Preview RigDisplays the selected reference proportions and chosen rig guides.
Create MocapBVH / Add to CollectionUses the last completed result downstream; changing options does not rewrite an already created BVH node.
Create T-pose BVH / Create A-pose BVHCreates separate single-frame reference rigs using the last skeleton; A-pose lowers the arms by 45 degrees.
Open result folderOpens the generated files.

Using this node

  1. Set up the body runtime and generate a short prompt.
  2. Preview the completed take and inspect the rig.
  3. Add it to a collection for retargeting/rendering; configure HandMDM separately if hand motion is needed.

Preview Rig also exposes finger bend/spread limits, tapered capsule end radii, and Collision preview (blocks rotations). Its Save action stores limits and radii in application settings. Use Rig Limiter to repair an existing generated clip without running generation again.

Hand articulation and finger/palm retargeting remain preview features. Changing generation hand settings requires another generation.

import

Load mesh or splat assets.

Inputs, parameters, buttons, and usage

Connections and result

Loads an asset file for the viewport or downstream geometry/splat processing. Use FBX Import or Mocap Import when you need their rig-specific validation.

Parameters and options

ControlWhat it does
File pathChoose the mesh or Gaussian-splat asset to load. Keep referenced files available when moving the workflow.
Asset summaryCheck the resolved file and asset information shown in the node before sending it to Scene.

Buttons and editing actions

ControlWhat it does
File and viewport controlsUse the path/file picker to choose the asset and the app’s viewport controls to inspect it. Import is a source node, not a conversion job.

Using this node

  1. Choose an existing asset file.
  2. Inspect its resolved path and preview.
  3. Connect it to Scene, Modeler, or the appropriate processing node.

fbx_import

Ingest and validate FBX rest geometry, capture pose, and animated pose roles for rigging workflows.

Inputs, parameters, buttons, and usage

Connections and result

The rest_geometry, capture_pose, and animated_pose roles describe the mesh/rest data, rig capture pose, and animation source. The validated result feeds retargeting, Scene, or rig-aware processing.

Parameters and options

ControlWhat it does
Rest geometrySource used for the rest mesh and skinning information.
Capture poseReference skeleton pose used to interpret the skin binding.
Animated poseSource of the animated rig; validation checks compatibility with the other roles.
Skin Weight ColorsColors the mesh by joint influence for skinning inspection.
Show Capture Joints / Show Animated JointsOverlays the corresponding skeletons on the mesh.
Debug loggingEnables additional diagnostic information when investigating an import.

Buttons and editing actions

ControlWhat it does
Validate FBX SourcesResolves the three roles and reports compatibility problems.
Setup FBX SDKInstalls or configures the runtime required for full FBX support.
ViewDisplays the validated import in the viewport.
Copy ReportCopies the complete validation report.

Using this node

  1. Provide the required FBX roles.
  2. Validate and resolve missing-runtime or incompatible-rig errors.
  3. View with the joint/weight overlays before retargeting or exporting.

fbx_animation

Import an FBX animation clip with skeleton validation and a viewport preview.

Use: Choose an FBX in path, click Validate FBX Anim, and inspect the clip and skeleton summary before previewing it.

Connect: Feed the node into fbx_import through animated_pose for animation or capture_pose for a capture source, or connect it to Scene for playback. The Create Node name is fbx_animation; fbx_animation_import is an alias.

Inputs, parameters, buttons, and usage

Connections and result

Reads skeleton animation from an FBX path. Use its animation as a retargeting source or inspect it independently.

Parameters and options

ControlWhat it does
PathFBX file containing the skeleton/animation to load.
Resolved pathShows the file actually used after resolution.
Clip and skeleton reportLists the parsed hierarchy and animation information; inspect it before choosing the take downstream.
Debug loggingAdds diagnostics for parsing and validation.

Buttons and editing actions

ControlWhat it does
Browse FBXChooses the source file.
Validate FBX AnimParses the skeleton hierarchy and animation clips.
ViewOpens the animated skeleton wire preview.
Copy ReportCopies validation details.

Using this node

  1. Browse to an animated FBX.
  2. Validate it and check its clip/skeleton report.
  3. View the animation before connecting it to retargeting.

genx_video_mocap

Run GEN-X/GEM-X video mocap and create downstream BVH import nodes.

Inputs, parameters, buttons, and usage

Connections and result

Processes source video through the GEN-X/GEM-X runtime and creates a downstream MocapBVH node from the generated result.

Parameters and options

ControlWhat it does
Source videoVideo file whose subject motion should be reconstructed.
Output folderRoot location for generated results.
Rig exportChoose SOMA BVH or Unitree G1 according to the downstream rig requirement.

Buttons and editing actions

ControlWhat it does
Browse controlsChoose the input video and output directory.
Open Full UIOpens the GEM-X Qt launcher for the complete runtime interface.
RunRuns reconstruction and creates a MocapBVH node when a BVH result is available.

Using this node

  1. Set up the GEN-X runtime and choose a source video.
  2. Choose the export rig and output folder.
  3. Run, then validate and preview the generated mocap before retargeting.

mocap_import

Import and validate BVH mocap clips with skeleton, frame, and clip summaries.

Inputs, parameters, buttons, and usage

Connections and result

Loads a BVH skeleton and motion clip for retargeting, collections, or viewport inspection.

Parameters and options

ControlWhat it does
PathThe input BVH file.
Resolved pathThe file used after path resolution.
Skeleton / frame / clip summaryInspect joint count, hierarchy, duration, and motion information after validation.
Debug loggingEnables additional diagnostics for the file reader.

Buttons and editing actions

ControlWhat it does
Browse BVHSelects a mocap file.
Validate BVHParses its hierarchy and motion data.
ViewShows the animated skeleton in the viewport.
Copy ReportCopies the validation report.

Using this node

  1. Browse and validate a BVH.
  2. View the motion and check its frame information.
  3. Connect the validated source to Anim Retarget or another animation consumer.

anim_retarget

Connect a mocap/FBX source to a target FBX rig with joint mapping, root controls, pose offsets, preview options, and rig bend/twist constraints. Preview and adjust the rig before exporting or using it as an animated cloth collider. Retargeting guide.

Inputs, parameters, buttons, and usage

Connections and result

Connect source and target rig contexts and the intended animation to the node’s matching ports. The output drives the target rig; validate both rigs before mapping.

Parameters and options

ControlWhat it does
Joint mapPairs source joints with target joints. Check left/right limbs, spine, and root correspondences.
Root source / Root target / Scale modeChoose the root pair and how source motion is scaled to the target; initial scale mode is auto.
Rotation modeControls how source joint rotations transfer; initial mode is copy_global_delta.
Pelvis constraintSelect a source/target pelvis pair independently of ordinary mappings. Snap matches position; Position Offset preserves the target’s positional offset while following aligned rotation.
Target pose offsetsAdjust the target reference pose joint by joint. These correct alignment before motion is applied.
Joint display size / curve thicknessChanges inspection handles and skeleton line visibility.
Animate Source and Target SkeletonsPreviews source motion and its retargeted target; disable to inspect reference poses.
Rig limits and collision guidesUse the rig preview controls to inspect and adjust bend/twist constraints before cloth or final output.

Buttons and editing actions

ControlWhat it does
Use Mapping RowSets the pelvis constraint pair from the selected mapping row.
Reset Joint / Reset AllClears pose-offset adjustments for one joint or the entire target.
ValidateChecks the source and target connections.
View / Copy ReportOpens the source/target preview or copies validation results.

Using this node

  1. Validate the source animation and target FBX rig.
  2. Map corresponding joints and align their reference poses.
  3. Preview several frames, then use the retargeted rig in Scene or as a cloth collider.

camera

Scene camera placement/control node.

Inputs, parameters, buttons, and usage

Connections and result

Connect the Camera to Scene. Its placement is edited in the Scene outliner/viewport, rather than camera-node position parameters.

Parameters and options

ControlWhat it does
FOVField of view in degrees; initial value 60. Wider values show more of the scene.
Aspect width / heightDefines the camera frame ratio; initial values 1920 × 1080. This is framing information, not a render-completion action.
Scene transformMove and rotate the camera in Scene to establish the shot.

Buttons and editing actions

ControlWhat it does
Scene camera controlsSelect and position the camera using Scene and viewport controls. There is no separate camera generation button.

Using this node

  1. Connect the Camera and visible assets to Scene.
  2. Choose the frame aspect and FOV.
  3. Position the camera and confirm the shot before configuring Render.

light

Directional, point, spot, and area light setup with intensity, range, and shadow parameters. The node picker also offers directional_light, point_light, spot_light, and area_light presets.

Inputs, parameters, buttons, and usage

Connections and result

Connect the light to Scene and use the Scene transform controls for placement and orientation. The node picker also provides directional, point, spot, and area presets.

Parameters and options

ControlWhat it does
Type / IntensityChoose the light type and brightness; initial intensity is 1.
RangeControls the light’s reach where applicable; initial value 300.
Width / HeightArea-light dimensions, initially 1 × 1.
Shadow strengthControls the darkness/contribution of the light’s shadow.
Shadow range / FOV / resolution / biasAdvanced shadow settings used by the relevant light/render path. Zero-valued overrides use the renderer’s applicable defaults.
PlacementPosition and rotation belong to the Scene outliner, not this node.

Buttons and editing actions

ControlWhat it does
Type and parameter editingEdit the light’s parameters and inspect the scene. This node does not require a separate bake/run action.

Using this node

  1. Choose a light preset and connect it to Scene.
  2. Place it and tune intensity/range.
  3. Adjust shadows only after the basic illumination and scene scale are correct.

instance

Create repeated scene instances.

Inputs, parameters, buttons, and usage

Connections and result

Connect a source mesh to mesh. The result supplies repeated instances to Scene.

Parameters and options

ControlWhat it does
CountNumber of copies; initial value 3.
PrefixOptional naming prefix for the generated instances.
MeshThe source geometry to repeat. Its resolved path is managed from the connection.

Buttons and editing actions

ControlWhat it does
Parameter editingChange Count and Prefix, then inspect the result in Scene; the node has no independent generation dialog.

Using this node

  1. Connect a source mesh.
  2. Set a small count and useful naming prefix.
  3. Add the result to Scene and arrange the instances as needed.

copy_to_points

Copy source geometry onto point inputs with instancing and material/texture support.

Inputs, parameters, buttons, and usage

Connections and result

Connect the placement geometry to points and the geometry to duplicate to copy.

Parameters and options

ControlWhat it does
Match NormalRotates each copy so its local up axis follows the target point normal.
GPU InstancesDraws the source mesh using per-point transforms. This is faster to render but does not create baked per-copy vertex ranges. Enabled initially.
PackIn baked mode, records each copied mesh range so later per-copy effects can deform it. It does not add those ranges to GPU instances.

Buttons and editing actions

ControlWhat it does
ViewDisplays the copies on the target points.

Using this node

  1. Connect target points and source geometry.
  2. Enable Match Normal if orientation should follow the surface.
  3. Use GPU Instances for drawing efficiency, or baked mode with Pack when later deformation needs per-copy ranges.

primitive

Generate procedural primitive geometry.

Inputs, parameters, buttons, and usage

Connections and result

Creates basic geometry that can feed Scene, Transforms, materials, or geometry-processing nodes.

Parameters and options

ControlWhat it does
ShapeChoose Sphere, Plane, Cube, Cone, Tube, or Torus.
Placement and scaleUse Transforms or Scene to position and resize the generated shape; the compact primitive selector is not a full modeling editor.

Buttons and editing actions

ControlWhat it does
ViewGenerates/resolves the selected primitive and previews it.

Using this node

  1. Choose a shape and View it.
  2. Connect it to Transforms or Scene.
  3. Apply materials or further geometry processing downstream.

curve

Create line, circle, or spiral curve scene assets.

Inputs, parameters, buttons, and usage

Connections and result

Creates a procedural curve asset for Scene and compatible point/curve workflows.

Parameters and options

ControlWhat it does
Curve typeChoose Line, Circle, or Spiral.
Generated pointsThe selected preset supplies the curve points and line segments used downstream. Use Scene transforms to place the curve.

Buttons and editing actions

ControlWhat it does
ViewDisplays the selected curve in the viewport.

Using this node

  1. Choose a curve type.
  2. View its point/line layout.
  3. Connect it to the curve-processing or placement workflow you need.

modeler

Inspect model object names, hide/isolate submeshes, and save object/point/edge/face selection groups.

Inputs, parameters, buttons, and usage

Connections and result

Connect mesh to inspect its submeshes and create named groups from object, point, edge, or face selections.

Parameters and options

ControlWhat it does
Viewport selection modeChoose the element type before selecting geometry; the group records that selection type and its element IDs.
Selection groupsStores named selections for reuse on the model.
Submesh visibilityEye controls show or hide objects in the outliner without deleting their source geometry.

Buttons and editing actions

ControlWhat it does
ViewDisplays the connected mesh for selection.
Group SelectionCreates a Modeler group from the active viewport selection.
Isolate SelectedRestricts the visible content to the selection.
Show AllRestores visibility for the model’s objects.

Using this node

  1. Connect a mesh and View.
  2. Choose point/edge/face/object mode and select elements.
  3. Group Selection to preserve the selection, then isolate it to check coverage.

transforms

Apply translate/rotate/scale offsets in-chain.

Inputs, parameters, buttons, and usage

Connections and result

Connect the incoming geometry to mesh. The result carries the selected transform or its baked geometry downstream.

Parameters and options

ControlWhat it does
PositionX/Y/Z translation; initial 0,0,0.
RotationX/Y/Z rotation; initial 0,0,0.
ScaleX/Y/Z scale factors; initial 1,1,1.
Auto bakeControls whether changes are baked automatically; initially disabled.
Source / pathIdentifies the resolved input and processed geometry path; the mesh connection supplies the source.

Buttons and editing actions

ControlWhat it does
ViewPreviews the transform on the connected asset.
BakeWrites the transformed geometry for downstream use.

Using this node

  1. Connect the mesh and adjust position, rotation, or scale.
  2. View to confirm axes and proportions.
  3. Bake when the next operation needs the transformed geometry rather than only its placement.

ply_sequence

Sequence playback/control for PLY streams and splat frame workflows.

Inputs, parameters, buttons, and usage

Connections and result

Connect a mesh/PLY source or choose the first frame manually. The node resolves a sequence for animated viewport playback.

Parameters and options

ControlWhat it does
SourceThe first PLY frame. A connected source takes precedence and disables manual browsing.
EnabledEnables sequence behavior.
LoopRepeats the sequence during playback; initially enabled.
Follow CamEnables the sequence’s follow-camera behavior; initially disabled.

Buttons and editing actions

ControlWhat it does
Choose fileSelects the first frame when no input drives the source.
Reload fileReloads the selected source sequence.
ViewOpens the sequence in the viewport.

Using this node

  1. Select the first frame of a consistently named PLY sequence.
  2. Reload and View it.
  3. Play the timeline and check Loop and Follow Cam for the intended shot.

image_gs_splat

Download a URL or local image, run Image-GS, preview the generated Gaussian result in the node, and output a viewport-ready PLY splat. Open detailed page.

Inputs, parameters, buttons, and usage

Connections and result

Provide an image URL or local image. The output is a viewport-ready Gaussian-splat PLY with generated preview/report files.

Parameters and options

ControlWhat it does
Generation modeExact Pixel exports a direct front-facing per-pixel splat sheet. Image-GS runs the optional optimization runtime.
Gaussians / Steps / Render heightControl optimization capacity, training duration, and image resolution. Increasing these can increase time and memory use.
Sheet scale / Radius scale / AlphaAdjust the sheet’s size, splat footprint, and opacity.
InitializationChoose gradient, saliency, or random initialization for optimization.
Auto HQEnables progressive error-guided training and quality settings based on source-image detail.
Quantize / PointsControls the available quantization and Gaussian visualization options.
Advanced quality parametersProgressive additions, loss ratios, axis limits, anisotropy, coverage, and exact-pixel limits refine output quality; use the detailed Image-GS guide for the full tuning context.

Buttons and editing actions

ControlWhat it does
Setup Image-GSInstalls or repairs the optimization runtime.
Browse / GenerateChooses an image and creates the result in the selected mode.
View / FolderPreviews the splat or opens its files.
StopStops active processing.
Debug report: Refresh / Copy / Open LogsUpdates, copies, or opens diagnostic output.

Using this node

  1. Start with Exact Pixel to confirm the input and scale.
  2. View the PLY in Scene.
  3. For optimization, set up Image-GS and try Auto HQ before manual advanced tuning.

A single-image splat sheet is not a reconstructed, fully observed 3D object. The dedicated Image-GS guide remains available from this catalog.

scene

Assemble and manage multi-asset scenes.

Inputs, parameters, buttons, and usage

Connections and result

Collects connected assets, cameras, lights, rigged characters, cloth, and effects. Connect Scene to Render or an export node for delivery. Rig Limiter and Animation Correction outputs can connect directly: Scene resolves their processed animation and displays the skeleton clip. Save Animation Correction edits before rebuilding Scene; use Anim Retarget first when you need an animated character mesh.

Parameters and options

ControlWhat it does
Outliner / visibilitySelect objects and use eye controls to show or hide them.
Position / Rotation / ScaleTransforms the selected scene entry; camera and light placement also lives here.
Hierarchy and skeleton rowsInspect object children and skeletons; select joints for the relevant rig controls.
Look At target / live toggleAssign an object or skeleton-joint target. Live follows viewport changes immediately; otherwise evaluation occurs on timeline frames and rebuilds.
Cloth arrangementMove, rotate, and scale individual panels before simulating. Placement is saved with the workflow and changing it resets the cloth.
Cloth collider rowsChoose the meshes/owners that interact with the cloth. Multiple colliders can be added.

Buttons and editing actions

ControlWhat it does
View SceneBuilds and displays the assembled scene.
Visibility controlsShows/hides selected scene entries.
Add / Remove colliderChanges the collider rows for a cloth object.
Browse skeleton jointsLets you drag a target joint or double-click to copy its target path.

Using this node

  1. Connect geometry, lights, and a camera, then View Scene.
  2. Arrange objects and confirm visibility and animation.
  3. Connect Scene to Render or an exporter and test a frame before a batch.

export_fbx

Export scene geometry to FBX.

Inputs, parameters, buttons, and usage

Connections and result

Connect Scene to scene. This exporter writes scene geometry to an FBX file; use Export FBX Animation for the rig/animation-specific workflow.

Parameters and options

ControlWhat it does
OutputDestination FBX file.
Include hiddenIncludes hidden scene entries when enabled; initial value is on.
Triangulate (Unreal compatible)Converts exported faces to triangles for the receiving application; initially off.

Buttons and editing actions

ControlWhat it does
BrowseChooses the destination path.
Export FBXWrites the connected scene geometry with the selected options.

Using this node

  1. Check the Scene and its visibility.
  2. Choose the output file and triangulation/hidden-object settings.
  3. Export and inspect the file in the receiving application.

export_fbx_animation

Export validated FBX rigs with skeletons, skinned meshes, animation ranges, FPS, and optional joint-scale curves.

Inputs, parameters, buttons, and usage

Connections and result

Connect a validated rig context to source. The exporter writes skeleton, supported skinned meshes, and optional sampled animation.

Parameters and options

ControlWhat it does
OutputDestination FBX file.
Export animationIncludes the animated clip when enabled.
Start / EndFrame range to export; initial range 0–120.
FPSSampling rate for the exported animation; initial value 30.
Scale curvesIncludes joint-scale animation; enabled initially. Disable only when the target does not support it.

Buttons and editing actions

ControlWhat it does
BrowseChooses the output file.
Export FBXExports the rig and selected animation range.

Using this node

  1. Preview and validate the rig or retarget result.
  2. Set FPS and frame range to match the intended clip.
  3. Export, then check skinning and motion in the destination application.

export_obj

Export to OBJ/MTL.

Inputs, parameters, buttons, and usage

Connections and result

Connect Scene to scene. Writes static geometry and related OBJ/MTL material information.

Parameters and options

ControlWhat it does
OutputDestination OBJ file.
Include hiddenIncludes hidden scene entries when enabled; initially on.
Triangulate facesConverts polygons to triangles during export; initially off.

Buttons and editing actions

ControlWhat it does
BrowseChooses the destination.
Export OBJWrites the geometry with the selected visibility and triangulation options.

Using this node

  1. Assemble and inspect the Scene.
  2. Choose an output path and export options.
  3. Export and keep associated material/texture files with the OBJ.

OBJ is a static geometry format; it does not preserve the skeleton animation workflow.

Mesh, Materials, and Grooming

Cloth Solver

cloth_solver

Simulate saved pattern panels with CPU or CUDA, sewing, elastic edges, stretch and bend controls, collisions, self collision, gravity, and pins.

Use and connections: Connect Pattern Design to pattern, an optional static/animated mesh to collider, and the result to Scene. Arrange panels, then Play or Step, or enable Simulate with timeline. Bake, Load cache, Save state, and Stop & save support reusable results and resuming a settled pose.

Cache behavior: Simulation-setting changes clear calculated timeline frames and unload outdated baked playback. Preview and rendering recalculate from the starting pose; saved drape inputs and files on disk are preserved.

Workflow guide →
Inputs, parameters, buttons, and usage

Connections and result

Connect a saved Pattern Design document to pattern, an optional static/animated mesh to collider, and cloth to Scene. Scene collider-owner rows can select additional collider meshes.

Parameters and options

ControlWhat it does
Enabled / BackendEnable the solver and choose auto, cpu, or cuda. CUDA requires a compatible configured runtime.
Stretch stiffness / Bend stiffnessResistance to stretching and folding. Higher stiffness generally needs adequate substeps/iterations to solve reliably.
Sew stiffness / Sew closure / Seam gapControls how strongly paired seams pull together, the closure duration in seconds, and the target separation in meters. A 1–2 second closure can pull seams together more gently around a body.
Surface density / DampingFabric mass per square meter and velocity damping. A density of 0.2 kg/m² is 200 g/m²; gravity acceleration is not reduced by choosing a lighter fabric.
Gravity / Gravity YEnable gravity independently of its stored strength. Earth gravity is −9.81 m/s².
Collision thickness / FrictionSurface clearance in meters and resistance to sliding. Try 0.003–0.005 m for body clearance; friction 0 is slippery.
Self collision / Ground plane Y = 0Enable fabric-to-fabric contact or collision with the ground plane.
Collider ownerDrop Scene meshes into the owner field or use comma-separated owner names. Leave empty to use the collider input.
Mesh spacing / Scene units per meterMesh spacing is in millimeters; smaller spacing creates a finer cloth mesh. Use 1 for meter-scale scenes or 100 for centimeter-scale scenes.
Substeps / Constraint iterations / Collision qualityMore solves/contact passes improve difficult motion at higher cost. The UI suggests 8 substeps for dressing, 16 for fast motion, and collision quality 4 as starting points.
FPS / Start frame / End frameSets standalone simulation timing and bake range. Timeline simulation uses the viewport timeline FPS.
Simulate with timelinePlaying or scrubbing simulates from the current or loaded starting pose and keeps calculated frames in memory. The cloth toggle beside FX changes this option for all solvers in the current view and stays synchronized with the node checkboxes. Toggling off/on preserves calculated frames; save the graph to persist the node settings.
Grab cloth with mouseAllows interactive cloth manipulation in the viewport.

Buttons and editing actions

ControlWhat it does
Play / Stop / StepRuns realtime settling or advances one solve step; in timeline mode, Step advances one animation frame and playback uses viewport controls.
ResetReturns the cloth to its starting arrangement.
Pin selected points / Unpin selected / Clear pinsUses viewport point selection to edit pins. Pin edits retain the current pose and recalculate subsequent frames.
Bake…Calculates the Start–End range from the arranged pattern and saves an animation cache.
Load cache… / Clear cacheLoads saved pose/animation data or unloads it and returns to the arranged pattern.
Save state…Saves pose, velocity, and sewing progress so simulation can resume later.
Stop & saveStops baking after the current solve and saves completed frames.

Using this node

  1. Save the pattern and arrange its panels around the collider in Scene.
  2. Use a short Play/Step test to check scale, seams, and collisions.
  3. Set the timeline/bake range and save a state or bake once the drape is suitable.

Changing simulation settings clears calculated timeline frames and unloads outdated baked playback. Preview/render recalculate from the starting pose; saved drape inputs and files on disk remain.

Pattern Design

pattern_design

Draft garment panels in an embedded canvas or Open Full Editor. Edit millimeter dimensions, points, cubic curves, linked mirrors, combined outlines, sewing pairs, and elastic edges.

Use and connections: Save a standalone .pattern.json document before using the pattern output. Closed, valid panels feed Cloth Solver or can be placed directly in Scene. The workflow references the pattern file; it does not embed unsaved edits or copy the pattern when saved.

Workflow guide →
Inputs, parameters, buttons, and usage

Connections and result

The pattern output references a standalone .pattern.json file. Connect it to Cloth Solver or Scene after saving valid closed panels.

Parameters and options

ControlWhat it does
Width / HeightDimensions for new rectangle/trapezoid panels, in millimeters; initial dimensions are 300 × 400 mm.
Selected edge / curve handlesChoose an edge and edit its geometry. Cubic edges expose handles for smooth shaping.
Point and panel selectionSelect points, edges, or panels for editing, moving, combining, or mirrored editing in the canvas/full editor.
Sewing pairsPairs panel edges so Cloth Solver can pull the corresponding boundaries together.
Elastic enabled / Strength / Rest lengthEnable contraction on selected edges. Higher strength pulls harder; 80% rest length shortens the desired edge length by 20%.
Linked mirrorsUse linked mirrored panels where symmetric edits should remain related.

Buttons and editing actions

ControlWhat it does
New Pattern / Open-LocateCreates a document or locates an existing pattern file.
Save .pattern.jsonWrites the standalone file used by the output.
Open Full EditorOpens the larger pattern editing workspace.
+ Rectangle / + TrapezoidAdds a panel from the dimension fields.
Draw Points / CloseDraws an outline and closes it into a panel.
Make Cubic Curve / Insert Point / Delete Point(s)Changes edge shape or point topology.
Sew / Elastic toolsSelects the edge-pairing or elastic editing operation.
Undo / Redo / FitReverts/reapplies edits or frames all panels.

Using this node

  1. Draw or add closed panels and set their dimensions.
  2. Pair seam edges and adjust curves or elastic settings.
  3. Save the pattern file, then arrange it in Scene and connect Cloth Solver.

Saving the workflow does not embed unsaved pattern edits or copy the pattern file. Keep the referenced .pattern.json with the project.

uv_unwrap

UV workflow stage.

Inputs, parameters, buttons, and usage

Connections and result

Connect geometry to mesh. The processed mesh supplies UV coordinates for texture, mask, and related workflows.

Parameters and options

ControlWhat it does
Mesh / sourceThe connected geometry to unwrap.
PathThe processed mesh path used downstream; check that the input has resolved before previewing.
UV inspectionInspect the result with a checker texture before relying on it for painting or grooming. This compact node does not expose an island-packing editor.

Buttons and editing actions

ControlWhat it does
ViewProcesses/resolves and displays the connected mesh for inspection.

Using this node

  1. Connect the mesh to UV Unwrap.
  2. View and apply a checker texture downstream.
  3. Inspect distortion before adding a painted mask or detailed texture.

normals

Smooth mesh normals and cache an updated OBJ for downstream texture, mask, and scene workflows.

Inputs, parameters, buttons, and usage

Connections and result

Connect mesh. The node computes smoothed normals and caches a processed OBJ for downstream geometry, texture, mask, or Scene use.

Parameters and options

ControlWhat it does
AngleSmoothing threshold in degrees; initial value 180 permits broad smoothing. Reduce it when hard surface boundaries should remain distinct.
Weld toleranceTolerance used when matching nearby vertices for smoothing; initial value 0.0001.
Source / pathThe original mesh and processed result path. Work at a consistent scale before changing the tolerance.

Buttons and editing actions

ControlWhat it does
ViewBuilds and previews the normal-adjusted result.

Using this node

  1. Connect a mesh with unwanted faceting or seams.
  2. Adjust Angle and test a small weld tolerance.
  3. View under lighting before passing the mesh to materials or Scene.

texture

Base texturing stage.

Inputs, parameters, buttons, and usage

Connections and result

Connect geometry to mesh and choose an image texture. The result can feed Scene or another material/texture stage.

Parameters and options

ControlWhat it does
TextureImage file applied to the connected mesh.
Source / pathResolved geometry from the mesh connection.
PreviewShows the chosen texture; the View control reports whether a mesh or texture is still missing.

Buttons and editing actions

ControlWhat it does
Choose textureBrowses for the image file.
ViewDisplays the textured model or connected scene.

Using this node

  1. Connect a mesh with suitable UVs.
  2. Choose the texture image.
  3. View and check orientation/coverage before rendering.

texture_pro

Expanded texture processing controls.

Inputs, parameters, buttons, and usage

Connections and result

Connect geometry to mesh. Generates a procedural Checkerboard or Matrix Rain texture for the model/scene.

Parameters and options

ControlWhat it does
Pattern / ResolutionChoose Checkerboard or Matrix Rain and texture resolution; available resolutions include 256, 512, and 1024.
Tiling / Pack X / Pack YControls tile density and horizontal/vertical packing.
Offset X / YShifts the checker pattern in cell units.
Speed / Direction / PanSets animation rate and rain direction. Pan slides cells smoothly; disabling it steps by cell.
Life min / maxMinimum and maximum lifetimes of rain chains, in seconds.
Chain gap min / maxTime interval range between new rain-chain spawns.
BG Color / BG AlphaBackground color and opacity.
Lighting / Emissive / SoftnessControls lighting contribution, procedural-text glow, and glyph-edge softness.

Buttons and editing actions

ControlWhat it does
BG ColorOpens the background color chooser.
ViewDisplays the procedurally textured model or scene.

Using this node

  1. Select the pattern and View on a UV-mapped mesh.
  2. Set tiling/packing before tuning animation.
  3. For Matrix Rain, adjust chain timing and glow while playing the timeline.

texture_layer

Layered texture compositing.

Inputs, parameters, buttons, and usage

Connections and result

Connect the target geometry to mesh and texture-producing sources to base and overlay.

Parameters and options

ControlWhat it does
BaseThe lower texture layer.
OverlayThe layer composited over the base; use the source layer’s transparency to control its contribution.
MeshGeometry receiving the composed texture.
Source layer settingsEdit texture generation, colors, and opacity on the upstream nodes; this compact compositor does not present a full layer-stack editor.

Buttons and editing actions

ControlWhat it does
ViewDisplays the combined layers on the connected model/scene.

Using this node

  1. Connect a mesh and a base texture.
  2. Connect the overlay texture and set its upstream alpha as needed.
  3. View to verify the combined result.

material

Material setup and conversion utilities.

Inputs, parameters, buttons, and usage

Connections and result

Connect mesh to apply the material settings before Scene or export.

Parameters and options

ControlWhat it does
TransparencyControls how much of the surface is transparent.
IOR / RefractionSets the index of refraction and refractive contribution used by the material.
Tint colorColors the material’s transmitted/tinted appearance.
Fresnel amount / colorControls the angle-dependent contribution near grazing views.
Source / pathGeometry receiving the material; the connection supplies the source.

Buttons and editing actions

ControlWhat it does
Color swatchesChoose tint and Fresnel colors.
ViewPreviews the material on the connected geometry.

Using this node

  1. Connect the mesh and View it with scene lighting.
  2. Set transparency and tint first.
  3. Adjust IOR/refraction and Fresnel while inspecting both front and grazing angles.

mask

Paint foreground/background mesh masks with brush, falloff, invert, clear, and preview controls.

Inputs, parameters, buttons, and usage

Connections and result

Connect mesh to paint a surface mask. Feed the mask result to Groom Guides or other mask-aware processing.

Parameters and options

ControlWhat it does
Foreground / BackgroundPaint and clear colors; initial values are white foreground and black background.
Brush sizeBrush footprint; initial value 48.
FalloffEdge softness of the brush; initial value 0.35.
Brush modeSelects the painting operation, initially add.
Mask pathStored mask result used downstream.

Buttons and editing actions

ControlWhat it does
Paint brushEnables mask painting in the 3D viewport.
InvertSwaps foreground and background colors.
ClearFills the mask with the background color.
ViewDisplays the mask on the connected model.

Using this node

  1. Connect a mesh and View.
  2. Set brush size/falloff and paint the intended region.
  3. Inspect boundaries before connecting the mask to Groom Guides.

split_volume / volume_selector

Filter/split geometry by volume.

Inputs, parameters, buttons, and usage

Connections and result

Connect the geometry to mesh and the selection volume to volume. Also registered as volume_selector.

Parameters and options

ControlWhat it does
VolumeDefines the spatial region used to filter/split the mesh.
InvertReverses which side of the volume selection is kept; initially off.
Source / result pathThe input geometry and processed result; inspect the split before later processing.
DebugEnables diagnostic output when investigating the result.

Buttons and editing actions

ControlWhat it does
ViewPreviews the connected selection/result.
SplitPerforms the volume-based geometry split.

Using this node

  1. Connect the mesh and volume.
  2. Place the volume and choose whether the selection is inverted.
  3. Split and inspect the result before export or downstream processing.

fx

Post-process and trail-style pipeline stage.

Inputs, parameters, buttons, and usage

Connections and result

Connect the animated source to mesh and optional repeated geometry to instance. This node builds trail-style effects from sampled motion.

Parameters and options

ControlWhat it does
Enabled / GlobalEnables the trail and selects global-space behavior.
Samples / Frame stepNumber and spacing of motion samples; initial values 28 and 1.
Spawn rate / Substeps / LifespanControls how frequently trail elements appear, intermediate sampling, and their lifetime.
Repeats / Radius / SidesControls repeated trail geometry and its thickness/resolution.
Color / Line widthSets trail appearance.
Profile rampShapes the trail profile along its extent.
Age scale min / max / rampChanges element size over its lifetime.

Buttons and editing actions

ControlWhat it does
WaveResets the profile ramp to its wave-shaped preset.
LinearResets the age-scale ramp to a linear progression.
Ramp editingAdjust the ramp points to control the shape and age response.

Using this node

  1. Connect an animated source and view it through Scene.
  2. Start with a short lifespan and modest sample count.
  3. Tune the profile and age scale while playing the animation.

fx_splat_physics

Add motion lag, gravity, noise, and fading trails to animated skinned splats.

Connect: Feed skinned_splat_proxy into splats, then send the result to Scene. Physics follows the animated target positions while allowing controlled drift.

Controls: Choose a preset, then tune Follow, Drag, Velocity, Gravity Y, noise, substeps, and Max Lag. Trail and Glow add visual effects; Reset on timeline jumps resets the simulation when scrubbing beyond the Jump threshold.

Inputs, parameters, buttons, and usage

Connections and result

Connect an animated Gaussian-splat source to splats. Adds follow/lag, force, trail, and physics-driven glow behavior during playback.

Parameters and options

ControlWhat it does
Enabled / PresetToggle physics and choose a starting preset. Apply copies the preset into the node values.
Follow / Drag / Velocity / ScaleControls attraction to the moving source, motion damping, inherited velocity, and overall physics scale.
Noise Type / Noise / N Scale / N SpeedChoose None, Curl Force, Velocity Multiply, Follow Multiply, or Drag Multiply. Set noise strength, pattern scale, and animation speed; lower N Scale makes broader noise, while higher values make tighter breakup.
Gravity YVertical force contribution; initial value −0.25.
Substeps / Max LagSubsteps improve temporal resolution; Max Lag limits displacement from the source.
Reset on timeline jumps / JumpResets history when a sufficiently large frame jump occurs; initial threshold 12 frames.
Trail / T Rate / T Life / T Alpha / T Size / T CurlT Rate is the fraction of source splats emitted per frame; T Life is their lifetime in frames. The other controls set starting opacity, size, and curl force.
Glow from physics / glow intensity / radius boostAdds a glow response and controls its strength and expanded splat footprint.

Buttons and editing actions

ControlWhat it does
ApplyApplies the selected preset to the current parameter values.

Using this node

  1. Connect an animated splat proxy and apply a mild preset.
  2. Play forward to inspect follow and drag.
  3. Add trails/glow after the core motion is stable, and enable jump resets when scrubbing.

fx_splat_fx

Apply audio-driven glow and traveling wave effects to Gaussian splats.

Connect: Supply a skinned splat proxy or a compatible splat-effects node to splats. Connect an audio source to audio or browse to an audio file, then use View to preview with the timeline.

Controls: Adjust audio gain, smoothing, threshold, and start offset, then tune glow intensity, radius, saturation, and wave strength, width, speed, and axis. This node is also available through the splat_fx and splat_glow aliases.

Inputs, parameters, buttons, and usage

Connections and result

Connect splats and an audio source to audio, or choose an audio file. Creates audio-reactive glow and waves on the splats.

Parameters and options

ControlWhat it does
Enabled / Audio pathEnables the effect and provides a local sound source when needed.
Analysis gain / Smoothing / ThresholdSets audio sensitivity, response smoothing in milliseconds, and the minimum activity level.
Audio start offsetAligns the audio analysis with the scene timeline in milliseconds.
Glow intensity / Radius boost / SaturationControls audio-reactive glow strength, footprint, and color saturation.
Wave strength / Width / Speed / AxisControls the deformation wave and chooses its X, Y, or Z direction.

Buttons and editing actions

ControlWhat it does
BrowseChooses the audio file.
ViewPreviews the audio-reactive splat effect.

Using this node

  1. Connect a splat source and audio.
  2. Align the start offset and tune gain/threshold.
  3. View during playback and adjust glow/wave strength.

colorize

Recolor splats with a three-color ramp driven by their shape or height.

Connect: Feed a skinned splat proxy or compatible splat-effects chain into splats, then preview it in Scene or with View.

Controls: Choose Thickness, Density, Radius, Area, or Height Y as the metric. Set low, middle, and high colors, then adjust percentile limits, invert, gamma, and blend. splat_colorize and fx_splat_colorize are aliases.

Inputs, parameters, buttons, and usage

Connections and result

Connect Gaussian splats to splats. Recolors them from a measurable splat property.

Parameters and options

ControlWhat it does
MetricChoose Thickness, Density, Radius, Area, or Height Y to drive the colors.
Low / Mid / High colorDefines the three-stop color ramp.
Low / High percentileBounds the mapped range while reducing the influence of extremes; initial values 2 and 98.
GammaAdjusts the distribution through the ramp; initial 1.
BlendMixes the recoloring with the source appearance; initial 1.
Invert ramp / EnabledReverses the mapping or toggles the effect.

Buttons and editing actions

ControlWhat it does
Color buttonsChooses low, mid, and high colors.
ViewPreviews the recolored splats.

Using this node

  1. Choose a metric appropriate to the shape.
  2. Set contrasting ramp colors and inspect the result.
  3. Tune percentile limits and Blend to preserve useful source detail.

fx_music_effects

Drive mesh displacement from an audio track, including animated character geometry.

Connect: Supply geometry to mesh and a sound source to audio, or select an audio file in the node. Use View to preview the deformation with timeline audio.

Controls: Set analysis gain, smoothing, threshold, and audio start offset. Displacement, maximum displacement, and outward-only movement control how far the surface responds to the sound.

Inputs, parameters, buttons, and usage

Connections and result

Connect mesh and audio or choose an audio file. Converts audio activity into mesh displacement.

Parameters and options

ControlWhat it does
Enabled / Audio pathEnables the effect and selects a file when the audio connection does not supply it.
Analysis gain / Smoothing / ThresholdControls sensitivity, temporal smoothing in milliseconds, and the minimum activity level.
Audio start offsetAligns the sound with the scene timeline in milliseconds.
Mesh displacement / Maximum displacementSets the displacement response and its upper limit.
Outward onlyRestricts deformation to outward motion; initially enabled.

Buttons and editing actions

ControlWhat it does
BrowseSelects an audio file.
ViewDisplays the mesh with the audio-reactive deformation.

Using this node

  1. Connect a mesh and a sound source.
  2. Align playback and start with a low displacement.
  3. Adjust gain and smoothing before raising the displacement limit.

wire

Wire and curve-based geometry path workflows.

Inputs, parameters, buttons, and usage

Connections and result

Connect mesh/point geometry to mesh. Creates line/curve paths using its point order and grouping attributes.

Parameters and options

ControlWhat it does
EnabledEnables the wire result.
OrderChoose id, file, x, y, z, probability, or hamming ordering along each path; initial value id.
Group attributeSeparates paths using an attribute such as name, the initial setting.
Source / pathThe input geometry and resolved result.

Buttons and editing actions

ControlWhat it does
ViewDisplays the generated wires/curves.

Using this node

  1. Connect geometry with the intended point order.
  2. Choose the grouping attribute so unrelated objects are separate.
  3. View and correct ordering before using the paths downstream.

groom_guides

Generate guide curves from a painted mask and source mesh.

Inputs, parameters, buttons, and usage

Connections and result

Connect a painted Mask to mask. The source mesh and mask define where guide roots are generated.

Parameters and options

ControlWhat it does
Guide countNumber of guides to seed; initial 256.
SegmentsSegments along each guide; initial 5.
ThresholdMask cutoff for eligible root locations; initial 0.35.
LengthInitial guide length; initial 0.08 in the source geometry’s scale.
SeedRepeatable random distribution of roots; initial 7.
Guides path / diagnosticsIdentifies the generated guide data and reports generation issues.

Buttons and editing actions

ControlWhat it does
ViewGenerates/resolves and previews the guides.
Diagnostic CopyCopies the report when a diagnostics dialog is opened.

Using this node

  1. Paint the growth region with Mask.
  2. Set guide density, length, and threshold.
  3. View, then connect the guides to Groom Deform or Guide Tube.

groom_deform

Bind guide curves to FBX/retarget rigs and optionally hide the deformer geometry.

Inputs, parameters, buttons, and usage

Connections and result

Connect guides and a compatible rig. Binds the guides to the animated character context.

Parameters and options

ControlWhat it does
ModeThe initial skinned_cv mode deforms guide control vertices with the rig.
Hide deformer geoHides the driving geometry in the result; initially enabled.
Rig compatibilityThe guides and rig must refer to compatible geometry/placement for roots to follow the intended surface.

Buttons and editing actions

ControlWhat it does
ViewPreviews the bound guides on the rig.
Diagnostic CopyCopies the available diagnostic report.

Using this node

  1. Generate guides on the character’s source mesh.
  2. Connect the corresponding animated rig.
  3. View several frames and check root attachment before adding simulation.

groom_guide_pose

Pose guide curves from animation frames with XPBD-style strand settings and optional GPU support.

Inputs, parameters, buttons, and usage

Connections and result

Connect guides and an optional collider. Produces a settled/simulated guide pose and a pose cache for later strand processing.

Parameters and options

ControlWhat it does
Enabled / Start frame / FPSEnables the strand calculation and sets its starting frame and timing; initial start 0 and FPS 24.
Substeps / IterationsTemporal subdivisions and constraint solves; initial values 4 and 8. Higher values cost more processing.
Stretch / Bend / DampingControls resistance to elongation and bending plus motion damping.
Gravity Y / Wind X-Y-ZForces applied to the guides; initial gravity −9.81 and wind 0.
Scene units per meterConverts scene scale for physical settings; initial value 100.
Root pinControls attachment strength at guide roots; initial 1.
Maximum velocityLimits motion velocity; initial 250.
DeviceBackend selection; auto initially. Choose a supported device for the configured runtime.
Settle seconds / Warmup framesControls how long the guide pose settles before being used. Initial warmup is 36 frames.

Buttons and editing actions

ControlWhat it does
Sim PoseCalculates the settled pose using the selected forces, timing, and collider.
Clear SimClears the stored simulated-pose result.
ViewDisplays the current guide pose.

Using this node

  1. Connect guides and the correctly scaled collider.
  2. Set the start frame and settling time.
  3. Sim Pose and inspect it; clear/recompute after changing the source setup.

groom_guide_sim

Simulate guides with frame, FPS, substep, stretch, bend, gravity, wind, and collider controls.

Inputs, parameters, buttons, and usage

Connections and result

Connect guides and an optional collider for animated strand simulation. The processed guide data can feed Guide Tube or Scene.

Parameters and options

ControlWhat it does
Enabled / Start frame / FPSEnables the strand calculation and sets its starting frame and timing; initial start 0 and FPS 24.
Substeps / IterationsTemporal subdivisions and constraint solves; initial values 4 and 8. Higher values cost more processing.
Stretch / Bend / DampingControls resistance to elongation and bending plus motion damping.
Gravity Y / Wind X-Y-ZForces applied to the guides; initial gravity −9.81 and wind 0.
Scene units per meterConverts scene scale for physical settings; initial value 100.
Root pinControls attachment strength at guide roots; initial 1.
Maximum velocityLimits motion velocity; initial 250.
DeviceBackend selection; auto initially. Choose a supported device for the configured runtime.

Buttons and editing actions

ControlWhat it does
ViewPreviews the guide simulation; use timeline playback to inspect the animated response.

Using this node

  1. Begin with properly attached or settled guides.
  2. Set scene scale, FPS, and collider.
  3. View and play forward, increasing substeps/iterations if the motion needs more stable constraints.

groom_guide_tube

Convert guide curves into tube geometry with radius ramp, subdivisions, caps, smoothing, and source-guide display.

Inputs, parameters, buttons, and usage

Connections and result

Connect guides to generate renderable tube geometry around the strands.

Parameters and options

ControlWhat it does
EnabledEnables tube conversion.
Root radius / Tip radiusThickness at each end; initial values 0.006 and 0.0015.
Radius profileRamp controlling thickness along the strand.
Sides / Segment subdivisionsCross-section resolution and along-strand subdivisions; initial 8 sides and 1 subdivision.
Caps / Smooth / GuidesControls closed ends, smoothed normals, and whether the source guides remain visible.

Buttons and editing actions

ControlWhat it does
FlatResets the editable radius profile to a flat profile.
ViewDisplays the tube geometry.

Using this node

  1. Connect posed or simulated guides.
  2. Choose root/tip radii and tune the radius ramp.
  3. View, then increase resolution only if the shot needs it.

skinned_volume_mesh

Generate skinned voxel collision meshes from rig contexts for grooming simulation and viewport collision.

Inputs, parameters, buttons, and usage

Connections and result

Connect a rig-aware mesh to mesh. Generates a skinned voxel-derived proxy suitable for collision/viewport use.

Parameters and options

ControlWhat it does
Density / Volume resolutionVoxel resolution along the longest model axis; higher values preserve more detail and require more resources.
Quality / Remesh smoothingNumber of smoothing passes. This rounds voxel edges; it does not add polygons.
Maximum influencesLimits skinning influences retained for the generated proxy.
Fill interiorControls interior filling of the volume.
Output directory / generated filesThe manifest, mesh, volume grid, and skin data describe the generated result.

Buttons and editing actions

ControlWhat it does
GenerateBuilds the proxy from the connected rig context.
View MeshDisplays the generated volume mesh.

Using this node

  1. Connect the validated rig/mesh.
  2. Generate at moderate density and inspect its silhouette.
  3. Increase density for missed detail or smoothing for blocky edges before using it as a collider.

skinned_splat_proxy

Generate skinned splat proxies from retargeted rigs, FBX imports, or guide tubes.

Inputs, parameters, buttons, and usage

Connections and result

Connect a supported rigged mesh, retargeted character, or guide-tube source to mesh. Generates sampled splats with skinning data.

Parameters and options

ControlWhat it does
Sample count / SeedControls point density and repeatable sampling; initial seed 1234.
Maximum influencesLimits the number of skinning influences retained per sample.
Color modeChooses the source of sample colors; initial mode skin_weights helps inspect attachment.
Radius scale / Adaptive sizeControls splat coverage and whether size adapts to the sampled geometry.
Hide source meshHides the original geometry when previewing; initially enabled.
Auto generate if missingAllows generation when no proxy exists; initially off.
Output directory / Manifest / PLY / Skin dataFiles making up the reusable proxy result.

Buttons and editing actions

ControlWhat it does
GenerateBuilds the splat samples and skinning data.
View ProxyDisplays the generated proxy.

Using this node

  1. Connect a validated animated source.
  2. Generate and View Proxy.
  3. Check coverage and deformation across several frames before applying splat effects.

Media and Review

html_preview

In-node HTML rendering with a path input and PDF export for generated decks and other upstream HTML artifacts.

Inputs, parameters, buttons, and usage

Connections and result

Connect an HTML file path to path, for example Task’s generated artifact. The document is rendered inside the node.

Parameters and options

ControlWhat it does
PathHTML file to display. A connected artifact path lets the preview follow generated output.
Document resourcesImages and styles referenced by the HTML must remain accessible beside the document or at their referenced locations.
Preview backendQt WebEngine provides the full browser preview and PDF export; a text fallback is used when it is unavailable.

Buttons and editing actions

ControlWhat it does
PDFExports the displayed HTML to a chosen PDF file. Generated deck layouts can supply their own page breaks.

Using this node

  1. Generate or choose an HTML document.
  2. Connect its path and review the rendered output.
  3. Use PDF when the document is ready to share.

PDF export requires Qt WebEngine; the text preview fallback cannot perform it.

qwen_image

Local Qwen Image 2.1 text-to-image generation and image editing with reference images, presets, seeds, and saved PNG results.

Generation and editing guide →
Inputs, parameters, buttons, and usage

Connections and result

Inputs are prompt, negative_prompt, image, and reference_image. Connected text or images override the corresponding local fields. A single-field Note supplies its value; a multi-field Note supplies checked values. An Image Collection connection supplies its first image path to the selected image input.

Outputs are image, image_path, seed_used, status, job_id, and metadata.

Parameters and options

ControlWhat it does
Backend and modelComfyUI offers INT8 and GGUF Q4_K_M. Native Diffusers uses a separate runtime and official BF16 weights; native INT8/GGUF inference is not validated and GPU output-quality validation remains in progress. Each backend requires its own setup.
Mode and promptChoose Text to Image or Image Edit. Enter the desired image or an editing instruction; Image Edit requires an input image.
ReferencesImage Edit accepts an optional second reference. Native also accepts additional reference paths, up to ten images total including the input.
Native Guidance and FramingGuidance can follow the prompt, favor the input composition, or use its pose/layout. Framing can use the last supplied image, match the input or first reference, or use custom dimensions. Guidance does not guarantee exact preservation.
Dimensions and presetsWidth/Height use a 32-pixel grid from 64 to 4096. Edit size controls edit conditioning and native automatic/matched pixel budget. Preview, quality, portrait, widescreen, and low-memory presets provide starting settings.
Steps, CFG, and negative promptAdjust inference settings. Negative prompts are native-only; native CFG 1 works without negative guidance, while CFG above 1 requires a negative prompt.
Seed and Random seedGenerate variations or reuse a fixed seed. The actual seed is available through seed_used.
Transparent PNG and Low memoryRequest transparency and choose a lower-memory execution profile. Inspect alpha warnings in the result. Native CPU offloading can be slower; restart ComfyUI when changing its launch memory profile.

Buttons and editing actions

ControlWhat it does
SetupChoose runtime, model storage where available, output folder, and permitted model use; read the license and install/verify the selected model.
Start Local Backend / Stop Local BackendStart the selected local runtime or stop this app's backend after active jobs finish or are canceled.
Check BackendCheck readiness for the selected runtime and model.
Generate / CancelGenerate the image or request cancellation of the active generation.
Open PNG / preview double-clickOpen the generated PNG in the system image viewer.
Output FolderOpen the save location. Change it in Setup.
Recover JobRetrieve the saved job's result after a timeout, restart, or detach without queuing a new generation.

Using this node

  1. Choose a backend, open Setup, review the model license and hardware/storage requirements, and install or verify the selected model.
  2. Start and check the backend, choose Text to Image or Image Edit, and enter a prompt and any required images.
  3. Choose a preset or inference settings and Generate. Inspect the preview, seed, status, and output PNG.
  4. Load saved variations into Image Collection for full-resolution review and pinned comparison.

Qwen's model-use acknowledgement does not grant commercial rights. Research/evaluation and separately licensed commercial use have distinct terms; read the linked Qwen Research License in Setup. Native generation does not silently fall back to ComfyUI.

image_collection

Arrange image sets, resize packed previews, and review original-resolution images with slideshow and pinned comparison controls.

Image collection and comparison guide →
Inputs, parameters, buttons, and usage

Connections and result

Stores ordered image paths for review and compatible consumers. Connect multiple Image Collections to the Design images input (reference_images) of MiniMax H3 Ref Video to combine character sheets, environments, and other designs. All collections contribute their paths, duplicates are removed in first-occurrence order, and the combined limit is nine images. Collections follow incoming source order, left-to-right then node name, while retaining their internal image order. A Qwen image or reference_image connection uses the first collection path.

Parameters and options

ControlWhat it does
Image pathsLoad adds the chosen files to the existing collection and removes duplicate path entries while retaining order.
Canvas previewSingle-image sizing follows the source aspect ratio. Multiple images use tightly packed rows, and a thin gray outline identifies the hovered image.
Move and resizeWith Edit off, drag an image to move the collection. Drag borders/corners to resize it; images reflow and load sharper previews as space grows. Node dimensions save with the graph. Source images are unchanged.
Supported file picker formatsPNG, JPG/JPEG, BMP, and TIF/TIFF are offered by the picker.

Buttons and editing actions

ControlWhat it does
Load / Explorer dropLoad appends files to the collection. Dropping local images into the 2D graph creates a new collection.
Edit / Remove / PathEnable Edit to select an image, copy its path, open its folder, or remove its entry. Removal does not delete the file. Turn Edit off to open the viewer or move the collection by dragging an image.
Double-click with Edit offOpen the clicked original in a large, resizable viewer with position, filename, and dimensions.
◀ ▶ / Left and Right keysLeft triangle or Left key: Previous. Right triangle or Right key: Next. Navigation wraps around. A pin changes navigation to the comparison sequence below.
Click / swipe in Fit modeClick for Next; drag left for Next or right for Previous. Fit mode uses the standard arrow cursor.
Wheel / mouse side buttonsWheel up goes Next; down goes Previous. Browser Back/Forward buttons go Previous/Next. These also work at 100% and follow pin behavior.
100%Show actual pixels. Clicking does not advance, and dragging uses the grab hand to pan. Turn it off to fit the image again.
Slideshow / Pause / SpaceToggle automatic Next every three seconds. Closing the viewer stops playback; Escape closes it. Navigation arrows and slideshow are disabled for a single image.
PinThe pin sits left of the navigation buttons: gray for an unpinned image, blue on the pinned reference. Previous shows the reference without stepping backward; Next restores the comparison, and a second Next advances. Click the blue pin to unpin or a gray pin on another image to replace it.

Using this node

  1. Load images or drop them into the 2D graph. Move or resize the collection; use Edit for path and removal controls.
  2. Turn Edit off and double-click an image. Browse, inspect at 100%, or play a slideshow.
  3. For comparison, pin A and go Next to B. Previous shows A; Next returns to B. Next again advances to C. Unpin to restore ordinary backward navigation.
  4. Connect the collection to a compatible image input as needed. Keep the referenced source files available.

Pin a reference for back-and-forth comparison

The pin button sits immediately left of the navigation triangles. It is gray on an unpinned image and blue when the displayed image is the pinned reference. Pin one reference to compare it against the other images without losing your place.

  1. Turn Edit off, double-click reference image A, and click its pin button. The pin turns blue.
  2. Go Next to comparison image B.
  3. Go Previous to show A. Repeated Previous actions stay on A; they do not step backward through the collection.
  4. Go Next once to return to B. Repeat Previous/Next to compare A and B as often as needed.
  5. From B, go Next again to advance to C. Previous now shows A, and Next returns to C. Forward browsing continues to wrap through the collection.

Pin A → Next: B → Previous: A → Next: B → Next again: C
Previous: A → Next: C

Click the blue pin on A to unpin and restore ordinary backward navigation. Click a gray pin on another image to replace the reference. The pin applies to the current viewer session.

The triangle buttons, Left/Right keys, mouse wheel, browser Back/Forward side buttons, Fit-mode clicks and swipes, and slideshow all use these same Previous/Next comparison rules. At 100%, clicking does not advance and dragging pans the image.

The workflow stores image paths, so keep the referenced files available when moving it.

video_player

Timeline playback and clip inspection.

Inputs, parameters, buttons, and usage

Connections and result

Connect a video or numbered image-sequence path to path, or choose a local file. For image-sequence delivery, connect Video Player to Sequence to MP4: the exporter receives the sequence path and can inherit this player's marked range.

Parameters and options

ControlWhat it does
PathThe media file to play. A connected path drives the field instead of the manual selection.
Scrub sliderSeeks through the full source. Cyan/yellow start/end flags mark the selected range; the bar outside them is darker in both the node and large player.
PreviewDisplays the current frame and reports when no usable media is loaded.

Buttons and editing actions

ControlWhat it does
BrowseChooses a local video.
Play / Pause / SpaceStarts or pauses playback, looping within the marked range. Space toggles playback for a selected Video Player when the graph has focus and throughout the large-player window. Space still enters text in the node's path field.
[ / ] / [X]Set the inclusive first/last frame at the playhead, or clear both flags with [X]. Image sequences use their filename frame numbers. Markers save with the workflow and reset when the source changes.
Large playerOpens original media in a maximized window with matching controls and minimize, maximize/restore, and close caption buttons. F11 toggles full screen; Esc returns to the framed window from full screen or closes a framed window. Closing returns the frame position to the node.
OpenOpens the loaded MP4 in the system's default external video player, like the MiniMax H3 Reference Video node's Open Video button. Available beside the large-player button in both the node and large-player window. Works with local or connected MP4 paths, including when the in-node preview is unavailable; disabled for image sequences, missing files, or an empty path.
Sound On / OffEnables or mutes audio for the preview.

Using this node

  1. Browse a clip or connect a generated mp4_path.
  2. Play and inspect its audio.
  3. Scrub and click [ and ] to mark the first and last frames; use [X] to clear them. Play, range, clear, and large-player controls sit on the left of the lower row; audio sits on the right.
  4. Open the large player to review the original-resolution source. For sequence export, connect this node to Sequence to MP4 and leave the desired boundaries on Auto.

youtube_downloader

Download a YouTube video as MP4, extract MP3 audio, and optionally save a Markdown transcript.

Use: Enter a video URL, choose an output folder, and enable transcript export with the preferred caption language when needed. Downloads and audio conversion use yt-dlp and FFmpeg; transcript availability depends on the video captions.

Outputs: mp4_path, mp3_path, and transcript_md. Connect the video to video_player for review or route the audio into the music and splat effects nodes.

Inputs, parameters, buttons, and usage

Connections and result

Connect a URL to url or enter it locally. Outputs mp4_path, optional mp3_path, and transcript_md for downstream review or analysis.

Parameters and options

ControlWhat it does
URLThe source video address; connected URL input takes precedence.
Output folderDirectory for the downloaded/generated files.
MP3Requests an extracted audio file; initially disabled.
Transcript / LanguageRequests a transcript and chooses its language; initial settings enable transcript with en.
Output paths / statusRead the generated file locations and progress/result messages.

Buttons and editing actions

ControlWhat it does
Choose output folderSets the download destination.
DownloadRuns the configured download and optional audio/transcript work.
OpenOpens the result/output location.

Using this node

  1. Enter the source URL and destination.
  2. Choose MP3 and transcript options.
  3. Download, then connect the resulting video/audio/text path to the appropriate consumer.

minimax_h3_video

Generate video from a prompt and optional first/last images through a local MiniMax H3 runtime.

Create Node listing: MiniMax H3 Local Video. For a connected Note prompt, a single visible parameter is automatic; with multiple parameters, only checked values are used. Labels are omitted, and an empty connected selection does not fall back to local text. The Note itself continues to expose all regular nonempty values.

Setup: Use Setup Model to prepare the local checkpoint, choose the model folder and variant, and configure a running local H3 service URL. Generation requires both the checkpoint and the local service.

Connect: Inputs are prompt, first_image, and last_image. Choose duration, aspect ratio, resolution, and output folder, then click Generate. Outputs are mp4_path, task_id, and h3_status.

Inputs, parameters, buttons, and usage

Connections and result

Connect prompt and optional first_image/last_image inputs. This node talks to a locally configured H3 generation service and exposes mp4_path, task_id, and h3_status.

Parameters and options

ControlWhat it does
Prompt / First frame / Last frameDefines the scene and optional image conditioning.
Model folder / VariantSelects the local model files and variant; the initial variant is fl2va.
Local URL / create-query-retrieve pathsConfigure the service endpoint. Initial paths are /generate, /status/{task_id}, and /files/{file_id}.
Duration / Ratio / ResolutionControls requested clip shape and length; initial duration 8 seconds, adaptive ratio, and 768p resolution.
Output folder / Poll interval / TimeoutChooses where the MP4 is saved, how often status is checked, and how long to wait. Initial polling is 5 seconds and timeout 1800 seconds.

Buttons and editing actions

ControlWhat it does
Setup ModelOpens the local model setup workflow.
Browse controlsChooses image, model, and output paths.
GenerateSubmits the request to the configured local service and retrieves its result.
Open / CopyOpens the generated video or copies its MP4 path.

Using this node

  1. Configure the local runtime/service and model folder.
  2. Test a short prompt with optional first/last images.
  3. Generate, monitor status, and review mp4_path with Video Player.

minimax_h3_api_video

Generate MiniMax H3 video through the cloud API using a prompt, frame images, or reference images.

Create Node listing: MiniMax API Vid. Note prompts follow the same selection rule as Qwen and the other H3 nodes: one visible parameter is used even unchecked; multiple parameters require checked values. Use MiniMax H3 Ref Video for previs-video, design-image, and audio references together.

Setup: Enter a MiniMax API key and select the API base, model, duration, aspect ratio, and resolution.

Connect: Use prompt with optional first_image and last_image in Image-to-video mode, or use reference_images in Reference-to-video mode. Reference images and first/last frames are separate modes. Outputs are mp4_path, task_id, and h3_api_status.

Inputs, parameters, buttons, and usage

Connections and result

Inputs are prompt, first_image, last_image, and reference_images. Outputs are mp4_path, task_id, and h3_api_status.

Parameters and options

ControlWhat it does
Prompt / ModeThe text prompt describes the video. Image mode uses first/last frames; Reference mode uses reference images and cannot mix them with first/last frames.
First / Last imagesAccept local supported images or the URL/data/file-reference formats shown by the node.
Reference imagesUp to nine references in the implemented H3 interface, including an Image Collection connection.
API key / Base URL / ModelConfigure the hosted service and selected model. The key must support the video service.
Duration / Resolution / RatioThe UI enforces combinations for its H3/H3-Max choices. Image-conditioned requests use adaptive framing; text/reference requests can use the offered ratio choices.
Create / Query / File pathsAdvanced service endpoint settings; query paths can contain {task_id}.
Output folder / Poll interval / TimeoutControls download location and request monitoring; initial poll interval 5 seconds and timeout 1800 seconds.

Buttons and editing actions

ControlWhat it does
Browse controlsChooses local conditioning images and output location.
GenerateSubmits the request, polls the task, and saves the resulting video.
Open / CopyOpens the MP4 or copies its path.

Using this node

  1. Choose Image or Reference mode and provide the matching inputs.
  2. Set credentials, model, and compatible clip settings.
  3. Generate and inspect the downloaded MP4 before using it downstream.

minimax_h3_reference_video

Generate a video guided by a 3D layout/previs clip, character sheets, environment designs, an optional audio reference, and a required prompt.

Create Node listing: MiniMax H3 Ref Video. The graph node keeps the full Reference Video name. This cloud API node is separate from MiniMax H3 Local Video and MiniMax API Vid.

Follow the previs and reference-video workflow →
Inputs, parameters, buttons, and usage

Connections and result

Connect prompt, reference_videos (Previs videos), reference_images (Design images), and optionally reference_audio. Browse local files or enter paths, public URLs, or mm_file:// references, one per line. Local media is inspected and uploaded automatically. Outputs are mp4_path, task_id, h3_reference_status, error, and metadata_path; connect mp4_path to Video Player.

Note prompts: With multiple visible Note parameters, only checked nonempty values enter the prompt, without labels. One visible parameter is used even unchecked. Selection also follows Note segments through Append/Switch. An empty connected selection reports an error; it cannot reuse local prompt text. This filtering happens in the H3 consumer, leaving Note output unchanged.

Parameters and options

ControlWhat it does
PromptRequired, up to 7,000 characters. Describe what each numbered reference should control: camera, timing, poses, character identity, environment, or lighting.
Previs videosUp to three MP4/MOV clips using H.264/HEVC, at most 50 MB each, 2–15 seconds per clip and 15 seconds total, at 23.976–60 FPS. Embedded audio must be AAC/MP3.
Design imagesConnect multiple Image Collections to this same pin. Their paths are combined in graph source order, each collection keeps its image order, and duplicate paths are removed. The combined limit is nine images, 30 MB each; supported formats are JPG/JPEG/PNG/WEBP/HEIC/HEIF.
Audio referencesUp to three WAV/MP3 clips, at most 15 MB each, 2–15 seconds per clip and 15 seconds total. Use at most 12 files across all reference types. Image/video dimensions must be 256–5760 pixels per side with a width/height ratio of 0.4–2.5.
Model / Duration / ResolutionMiniMax-H3: 4–15 seconds at 768P or 2K. MiniMax-H3-Max: 5–15 seconds at 480P or 768P. Both accept reference images/video/audio. Ratio can be adaptive or an offered fixed ratio.
H3 Max promptH3 Max prompt expansion: disabled, balanced, or quality.
API key / Save folderRequires an eligible MiniMax pay-as-you-go key. Enter it for this app session or set MINIMAX_API_KEY / MINIMAX_PAYG_API_KEY. Entered keys are excluded from saved workflows and task metadata. Outputs default to the workflow's videos/minimax_h3_reference folder, or temporary storage for an unsaved workflow.

Buttons and editing actions

ControlWhat it does
Use previs prompt template / BrowseInsert a camera/motion/design instruction template or append local reference files. Reference numbers follow list order within each media type.
Generate / Generate newUpload local references, submit a new generation, poll in the background, and download its MP4. A new generation uses provider balance; reference-video input is also billed.
Resume taskQuery and download the saved task without uploading the original inputs or submitting another generation. Save the workflow to retain the task ID across app restarts. Tasks are queryable for seven days.
Stop waitingStops local waiting after an in-flight request finishes or times out. Submitted work continues at MiniMax and may still incur charges; Resume retrieves the same task later.
Open videoOpens the downloaded MP4 in the system player.

Using this node

  1. Supply a short previs clip and connect character/environment Image Collections to Design images.
  2. Write the prompt or connect a Note, checking the desired values when it has multiple parameters.
  3. Choose the model and output settings, supply a pay-as-you-go key, generate, and review the MP4.

Reference following is generative: exact camera, framing, timing, or pose preservation is not guaranteed. This node uses reference roles; first/last-frame inputs belong to the separate frame-generation node.

wan22_video

Run local Wan 2.2 TI2V-5B text-to-video or image-to-video generation and output an MP4.

Setup: Use Setup Wan and select the runtime and model folders. Connect prompt and optionally image, or fill in those fields directly.

Controls: Set a negative prompt, output size, frame count, seed, sampler, steps, guidance, and Image Change. CPU offload and text-encoder controls help manage GPU memory. Outputs are mp4_path and wan_status; connect the MP4 to Video Player for review.

Inputs, parameters, buttons, and usage

Connections and result

Connect prompt and optional image. Runs the configured local Wan2.2 TI2V-5B runtime and exposes mp4_path and wan_status.

Parameters and options

ControlWhat it does
Prompt / NegativeDescribes what to generate and what to avoid.
ImageOptional starting image for image-to-video.
Runtime / ModelLocations of the Wan2.2 code/runtime and TI2V-5B weights.
Size / FramesThe size key sets target pixel area for image-to-video, which follows the input aspect ratio. Shorter clips generally drift less; initial frame count is 121.
Solver / SeedChoose the sampler, initially UniPC. A numeric seed allows repeatable runs; blank/random chooses a new seed.
Image changePreservation helper: lower values reduce effective Denoise and CFG for image-to-video; 1 is normal behavior.
Denoise / CFG / StepsDenoise is Wan sample shift; lower values can reduce warping. CFG controls prompt guidance. More steps take longer and do not necessarily fix deformation.
Offload / T5 CPU / DtypeReduces GPU memory use by offloading model parts, running the text encoder on CPU, or converting model dtype. These are initially enabled.
Output folderDirectory for generated MP4 files.

Buttons and editing actions

ControlWhat it does
Setup WanInstalls or repairs the local runtime and model.
Browse controlsChooses image, runtime, model, and output paths.
Generate / StopStarts or stops the local generation job.
Open / CopyOpens the result or copies its MP4 path.

Using this node

  1. Set up the runtime and weights, then choose an input image if needed.
  2. Test a short clip with memory-saving options enabled.
  3. Review the output and tune Image change/CFG before increasing clip length or step count.

render

Render sequence/export pipeline stage.

Inputs, parameters, buttons, and usage

Connections and result

Connect Scene to scene. The result can feed Post Process, Sequence to MP4, or For Each End’s render input.

Parameters and options

ControlWhat it does
CameraChooses the scene camera used for output; confirm it exists and frames the intended objects.
FPSOutput playback rate, initially 30. Scene animation and cloth are evaluated using the Scene timeline FPS.
FormatChoose PNG, JPG, TIFF, or EXR.
Start / EndFrame range. End = Auto uses the Scene end marker or animation length when no marker is set.
Output pathDestination for the rendered frame sequence.
Take folders / Take numberWrites separate take directories when enabled. For Each supplies its own take folder/number during a batch.

Buttons and editing actions

ControlWhat it does
BrowseChooses the output path.
Render SequenceRenders the configured range from the connected scene.
Stop RenderStops the active render from its progress interface.

Using this node

  1. View the Scene and select its camera.
  2. Render a small range to check framing and simulation.
  3. Render the full sequence or connect the completion output to For Each End.

Changing Render FPS does not change the Scene simulation step rate; set timeline FPS appropriately before cloth or animation work.

post_process

Apply image-sequence effects such as ordered dithering and produce an output frame pattern.

Inputs, parameters, buttons, and usage

Connections and result

Connect a sequence-producing source such as Render to source, or supply a supported source path. The output_pattern identifies the processed frame sequence.

Parameters and options

ControlWhat it does
EffectThe implemented processing action is ordered dithering.
LevelsNumber of quantization levels; initial value 4.
MatrixChoose a 2×2, 4×4, or 8×8 ordered-dither matrix.
StrengthEffect strength from 0 to 200; 100 is normal strength.
GrayEnables grayscale processing.
Output directory / Output pattern / Frame countChoose where processed frames are written and inspect the resulting pattern/count.

Buttons and editing actions

ControlWhat it does
Choose output folderSets the processed sequence destination.
Apply DitherProcesses the input sequence using the selected settings.

Using this node

  1. Connect a rendered sequence.
  2. Choose output folder, levels, matrix, and strength.
  3. Apply Dither and review frames before connecting Sequence to MP4.

sequence_to_mp4

Encode image sequences to MP4 with FPS, codec, and bitrate settings.

Inputs, parameters, buttons, and usage

Connections and result

Connect an image sequence from Render, Post Process, or Video Player to source, or provide a folder, numbered frame, or sequence pattern. Video Player supplies its image-sequence path and optional markers. This exporter encodes numbered images; it does not trim an existing MP4.

Parameters and options

ControlWhat it does
SourceInput sequence to encode; use the upstream output pattern where available.
OutputDestination MP4 path.
CodecChoose H.264 or H.265; initial selection H.264.
Start / EndInclusive filename frame numbers. Each boundary independently defaults to Auto, using the corresponding connected Video Player marker or the first/last source frame when no marker is set. Type a number to override that boundary only. For example, 1020–1040 exports 21 frames. Invalid, out-of-source, or missing-frame ranges are rejected.
FPSPlayback rate assigned to the frame sequence; initial value 24. Match the intended render playback rate.
BitrateTarget encoding bitrate, initially 12M. Higher values increase output size and can retain more detail.

Buttons and editing actions

ControlWhat it does
Choose output MP4 pathSets the destination file.
Convert to MP4Encodes the sequence through the configured ffmpeg support.

Using this node

  1. Connect a completed render or processed frame sequence.
  2. Set the intended FPS, codec, output file, and range. Leave Start/End on Auto to inherit Video Player flags, including through Post Process, or enter explicit source frame numbers.
  3. Convert and review the resulting MP4 in Video Player.