About this project

FireRed-OpenStoryline turns complex video creation into natural, intuitive conversations. The system combines smart media search and organization (automatically finding and downloading matching images/video clips), intelligent script generation with few-shot style transfer for specific tones (product reviews, vlogs, etc.), intelligent music/voiceover/font recommendations with beat-syncing, conversational refinement for rapid edits via natural language, and editing skill archiving to save and replay complete workflows as reusable Skills. Recent additions include AI transition generation (creating transition shots from frame descriptions), ASR-based rough cut skill for speech videos (removing filler words and disfluencies), and integration with OpenClaw and Claude Code via Agent Skills for installation and usage automation. The architecture uses a Model Context Protocol (MCP) server, video processing nodes, an agent skills library, and agent memory storage. It builds on MoviePy, FFmpeg, and LangChain. Installation supports Linux/macOS (automated via build_env.sh), Windows (manual resource download), and Docker. The web interface runs on FastAPI, with a CLI alternative. Configuration requires API keys in config.toml. Demos showcase multiple styles: Zhongcao, Humorous, Product Picks, Artistic, Unboxing, Talking Pet, Travel Vlog, and Year-in-Review. The project is licensed under Apache 2.0 and actively developed with a roadmap including voiceover editing, voice cloning, more effects, generative capabilities, and GPU acceleration.