Orkas-AI/Orkas-VideoStudio
Turn your coding agent into a video studio: describe a video in plain language, and your agent writes the timeline and produces the file.
What it solves
OrkasVideoStudio allows coding agents (like Claude Code, Cursor, or Codex) to compose, generate, and edit videos using natural language. It replaces "black-box" AI video generation with a transparent, editable, and diffable intermediate representation (plan.json), enabling precise control over video production without requiring the user to manually write code or use a proprietary GUI editor.
How it works
The system acts as a bridge between an AI agent and various video production tools. It provides the agent with the necessary knowledge (via skills) and deterministic capabilities through a CLI (ovs) and an MCP server. It operates across four production lines:
- Compose: Creates motion graphics and explainers using HTML and HyperFrames.
- Edit: Handles cutting, trimming, silence removal, and subtitle burning using
ffmpegandwhisper.cpp. - Generate: Integrates with external AI providers (OpenAI, Gemini, Doubao) via BYO keys for cinematic footage and talking heads.
- Auto: An end-to-end pipeline that coordinates the other three lines via a single
plan.jsonto build and verify the final cut.
Who it’s for
It is designed for developers and users of coding agents who want to automate video creation—such as creating social media explainers, cleaning up webcam recordings, or generating promotional clips—while maintaining the ability to audit and edit the production plan.
Highlights
- Agent-Agnostic: Works with any agent that can run a shell or speak MCP.
- Editable IR: Uses a
plan.jsonfile so that changes to a single line can trigger a re-render of only that specific segment. - Zero-Key Trunk: Core composition and editing features work without paid API keys; generation is opt-in and BYO.
- Delivery Guard: Includes a deterministic check (
promise-check) to verify the finished video meets the original requirements before shipping.