Summary

OrkasVideoStudio transforms coding agents into video production studios, enabling users to describe videos in plain language. It provides an end-to-end pipeline for video composition, editing, and AI-driven generation, outputting an editable, diffable timeline (plan.json) for precise control and re-rendering. This empowers agents and users with deterministic video creation capabilities.

How It Works

The system translates natural language prompts into a structured, editable plan.json file, acting as a video timeline and contract. This IR is processed by thin wrappers around HyperFrames, ffmpeg, and whisper.cpp. The approach is advantageous due to its agent-agnostic nature, deterministic execution, and the ability to re-render individual segments of the video by editing the plan.json, facilitating iterative development and precise control.

Quick Start & Requirements

• Installation: Currently requires cloning the repository and running pnpm install && pnpm build. npm package installation is planned.
• Prerequisites: Node.js (≥ 22) and ffmpeg/ffprobe on the system PATH are mandatory. AI generation features (image, video, TTS) require users to supply their own API keys (e.g., OpenAI, Gemini).
• Integration: Driven via the ovs CLI, MCP typed tools, or native skills for agents like Claude Code and Codex.
• Links: orkas.ai, github.com/Orkas-AI.

Highlighted Details

• Four Production Lines: Offers Compose (script to HTML motion graphics), Edit (real footage manipulation), Generate (AI footage via BYO keys), and Auto (end-to-end pipeline).
• Editable plan.json: A core feature providing a readable, diffable, and per-segment re-renderable video contract.
• Zero-Key Trunk: Core compose, edit, and transcribe functionalities operate without requiring user API keys.
• Agent Agnostic: Designed for seamless integration with any coding agent capable of shell execution or MCP communication.

Maintenance & Community

This project is part of the Orkas-AI open-source initiative. Further community or roadmap details are not explicitly detailed in the README.

Licensing & Compatibility

• License: MIT.
• Compatibility: Rendering relies on Apache-2.0 licensed HyperFrames 0.7.60. Editing uses system ffmpeg. Transcription uses HyperFrames/whisper.cpp. Generation features are opt-in and require BYO keys, with no managed backend or vendor lock-in.

Limitations & Caveats

The npm package installation is not yet available, necessitating source builds. AI generation capabilities are opt-in and depend on users providing and managing their own API keys and associated costs. The toolkit is derived from a larger application, suggesting potential for ongoing development and feature evolution.