OpenShorts

OpenShorts is an open source video platform for turning long recordings into vertical clips for TikTok, Instagram Reels and YouTube Shorts. It detects promising moments, follows faces, chooses layouts for one or two speakers and burns styled subtitles into the finished video. Podcasts, interviews, webinars, screencasts and livestream recordings are its main source material.

The clipping pipeline combines faster Whisper, MediaPipe, YOLO and FFmpeg. Transcription, tracking and rendering can run on your server, while an OpenAI compatible local model such as Ollama can handle moment selection. Optional Gemini analysis can help choose layouts and process silent or visual material. CPU processing works, but NVIDIA acceleration substantially reduces rendering time.

OpenShorts also includes AI actor videos, voice dubbing, thumbnail generation, YouTube metadata, scheduled social publishing and a REST API and MCP server. These features introduce external dependencies. AI actor generation uses fal.ai, voice work uses ElevenLabs and social publishing uses Upload Post. Relevant prompts, images, audio or credentials may reach those providers when you enable their integrations.

The application is MIT licensed and free to self host without watermarks or usage limits. A paid hosted service is available, making the overall pricing model freemium. Docker Compose starts the frontend, backend and renderer, but the supplied configuration exposes their ports and the self hosted API does not require authentication by default. Keep it on a trusted network or add authentication, HTTPS and a reverse proxy before remote access. Public gallery and S3 settings also need review before processing private client material.