edualnog/medusa-clip

Medusa Clip is a desktop application that takes long gaming video recordings and automatically finds the best moments, cutting them into short vertical clips in 9:16 format suitable for TikTok, Reels, and YouTube Shorts. The cutting engine is built specifically for gameplay: it reads audio levels, motion in the frame, and visual content to identify highlights such as clutch plays, failures, climactic moments, and reactions. The app runs entirely on your own computer. The video never leaves your machine, there is no account or login required, and the AI processing uses an API key you supply yourself from providers such as OpenAI, Anthropic, or OpenRouter. You pay the provider directly, and the README estimates the cost at a few cents per batch of clips for a ten-minute video. Output includes two reframing options: a facecam-plus-gameplay layout with a blurred background, or a full-gameplay focus layout. Each clip gets karaoke-style word-by-word captions and a hook headline automatically. The processing pipeline runs transcription on-device using GPU acceleration when available, then sends multimodal analysis to the AI provider to select and judge the best moments in parallel. The application is packaged as a self-contained installer for macOS, Windows, and Linux. The installer includes the Python processing engine, FFmpeg, and FFprobe, so users do not need to install Python or any other tools separately. The installers are not yet code-signed, so first-time users on macOS and Windows need to approve the app through an extra confirmation step. The README and project are written in Brazilian Portuguese, though the application itself functions in any language its transcription model supports. The project is free and open source under the AGPL-3.0 license. Contributions are welcome and the project has a Discord community.