Node reference — overview

Every node on the canvas is documented in this section. This page gives you the shared vocabulary.

Categories

Category	Nodes	Purpose
Input	Text Prompt, Image, Video, Audio	Bring text and media into the graph.
Model	LLM, Image Gen, Collection, Video Gen, TTS, Voice Changer, Video Analyzer, Ad Recast, Upscaler, Character Swap, Provenance Check	Call generative models and media tools.
Cinematic	Brief Interpreter, Screenwriter, Shot List Director	The curated writing flow for campaign work.
Template	Template Input, Template Output	Declare a template’s inputs and deliverables.
Output	Output	Pin a graph’s final result.

Port types

Ports are typed, and edges only connect matching types:

Type	Carries
text	Prompts, scripts, briefs — any string.
image	A single image (generated or uploaded).
video	A video clip.
audio	An audio clip.

Nodes that accept multiple reference images expose numbered slots (image_1 … image_10) — one connection each.

Parameters

Parameters are the controls on the node card. Conventions that hold everywhere:

• model — a dropdown over the curated model catalog. The catalog is maintained centrally, so new models appear without any change to your graph.
• Model-aware controls — a node only shows the parameters its selected model supports. If you don’t see a knob (e.g. temperature on some image models), that model doesn’t take it.
• prompt_enhancer — on prompt-driven nodes: off (default), generic, or cinematic. When enabled, your prompt is expanded by an LLM before the model sees it.
• Local text wins over wired text. Generation nodes take their prompt from a wired connection or their own field — and when both are set, the node’s own text takes precedence and the card shows an override banner. Clear the local field to let the wired value drive.