videofree — local AI video creation platform

Connect people, agents, and ComfyUI into one creative pipeline.

VideoFree is an AI creation hub that runs on your own computer: people create in the workspace, agents call in over MCP, and ComfyUI executes underneath — all three sharing one set of capabilities. Generation, the editor, even canvas nodes are plugins: register once and both humans and AI can use them, straight from plain language to final cut.

$ Download Read the docs

Local models
Cloud generation
Agent collaboration
Plugin distribution

agent_prompt — paste to Claude Code / Codex / Cursor

Install VideoFree on this computer: download the latest release from https://github.com/linpengishero/VideoFree/releases and extract it to [folder] (or clone the repo), follow README.md for my OS, then start it — double-click videofree.bat on Windows, or ./videofree.sh on Linux. Open http://127.0.0.1:8765 to confirm the workspace is up, then tell me what is still missing: the ComfyUI address, the plugins to install, and the models each plugin INSTALL.md declares (fetch them from Hugging Face or the official source). If a download needs a login, terms acceptance, or payment, stop and let me handle it.

Join the community Plugin sharing, generation help, and development talk

Discord community →

Why VideoFree

Local AI is powerful — and still hard to use

Cloud products pick the models and flow for you; going local turns you into a workflow maintainer. VideoFree adds a stable product layer on top of ComfyUI — from install to final cut, with zero dependencies to start.

Workflows as plugins

Turn a ComfyUI workflow you have debugged into a plugin: models, node graphs, parameter UI, and assets live in one plugin folder — copy it into plugins/, restart the gateway, done. The moment it registers, it is an MCP tool: users pick the capability in the workspace, agents call it directly.

core — plugin.json

Plain-language creation

Each plugin carries collaboration skills that ask, refine, and compile natural language into the execution prompt the model actually needs.

Local + cloud, one gateway

Local model plugins and cloud API plugins run under one gateway and one project — switch per shot, never locked to a single source.

Project-based management

Project → category → asset: scripts, prompts, jobs, and reusable assets all save with the project; final media auto-archives, and regenerating restores the exact prompt and parameters.

MCP: any AI can call it

MCP ships with the gateway — paste one URL and Claude Code, Codex, or Doubao can discover capabilities, launch generations, and manage projects. 30+ tools cover the whole workspace, and every registered plugin lands in the tool surface with zero MCP-side config.

Timeline editing & final cut

A built-in editor: five tracks — main, overlay, audio, captions, text — with transitions, waveform mixing, and one-click transcription; FFmpeg renders the final cut in-project.

Infinite Canvas: wire capabilities into pipelines

Image, video, speech, agents, and resource processing become canvas nodes — connect to compose, run with one click, resume after failures. Nodes are plugins; existing capabilities plug in unchanged.

in development Explore Infinite Canvas →

Plugin platform

The demo plugins are just the start — you define the capabilities

The repository ships demo plugins across image, video, speech, music, and post-production, showing what one plugin can package. Turning your own debugged ComfyUI workflow into a plugin is the real VideoFree workflow.

krea2 · qwen-image · gpt-image · minimax-h3 · ltx-2.5 · cosyvoice3 · qwen-tts · ace-step · minimax-music · video-post · film-studio · image-split · project-archive · ai-project-progress

See the demos and start building →

Docs

From install to full automation

Eight bilingual tutorials: install and use plugins, connect over MCP, then take the demo skill apart and build your own fully automated video pipeline.

01 /