jtydhr88/ComfyTV

What it solves

ComfyTV transforms ComfyUI from a standard node-based graph into a comprehensive, canvas-based media workbench. It solves the fragmentation of the creative process by integrating generation, editing, compositing, and exporting for images, video, audio, music, and 3D assets into a single unified flow, reducing the need to jump between different software for post-production.

How it works

ComfyTV operates as a custom node suite for ComfyUI, introducing a project-centric approach where each stage (node) runs independently. Instead of a global queue, downstream nodes consume snapshots of previous outputs, allowing users to iterate on specific parts of a chain without re-running the entire workflow. It integrates a wide array of specialized editors (layer editors, piano rolls, 3D viewports) directly into the nodes and provides a "Bridge" system to connect third-party ComfyUI plugins into its pipeline.

Who it’s for

It is designed for AI artists, motion designers, and multimedia creators who use ComfyUI and want a professional-grade production environment for complex generative media pipelines involving multi-modal assets.

Highlights

• Multi-modal Workbench: Over 190 stages covering image generation, 2D layer editing, professional video editing/compositing, audio processing, symbolic music composition, and 3D scene management.
• Agent-Native Integration: Includes an embedded chat bot, an MCP server with 45 tools for AI agents (like Claude Code), and a system for installable "Agent Skills" to automate production methodologies.
• Project-Centric Workflow: Maintains full history and assets within projects, featuring a dedicated asset library and integration with the Eagle app for archiving and provenance tracking.
• Rich In-Node Editors: Embeds functional tools like a PSD-compatible layer editor (Pentrado), storyboard workbenches, and live-preview video effects.
• Flexible Infrastructure: Supports remote ComfyUI runners, custom JSON workflow imports, and an FX Chain system to stack multiple effects for single-pass rendering.