OpenShorts

OpenShorts turns podcasts, interviews and other long videos into short clips for TikTok, Instagram Reels and YouTube Shorts. It's for creators and social media teams who want AI to select excerpts and prepare them for posting.

• Automatic clip selection: Choose Gemini or a local OpenAI-compatible server to score candidate moments from transcripts and prepare clip titles, descriptions and hook text.
• Vertical framing and captions: Face tracking keeps speakers in view, split layouts stack two speakers, and word-level subtitles appear directly on the video.
• Dubbing and AI actors: ElevenLabs can dub clips while preserving the speaker's voice; AI Shorts generates marketing videos with actors, lip-sync, voiceovers and supporting footage through external services.
• YouTube tools and publishing: Generate thumbnails, titles and descriptions with chapters, then publish or schedule videos for YouTube, TikTok and Instagram through Upload-Post.
• Agent automation: An MCP server, API and CLI let Claude, ChatGPT and n8n trigger the clipping workflow without using the dashboard.

The core application is MIT-licensed and builds on kamilstanuch/Autocrop-vertical. Self-hosting uses Docker with at least 8 GB of RAM. It runs on a CPU and can use an NVIDIA GPU for acceleration. Optional hosted-service and billing components have separate commercial terms and aren't required to self-host the core.

Local clipping can use Ollama, LM Studio, vLLM or another OpenAI-compatible server, keeping transcript-based moment selection and video processing on your infrastructure. Automatic layout selection, on-screen content detection and silent-video analysis still need Gemini. Without its key, speaking videos fall back to face-tracking crops and silent videos can't be processed. Dubbing uses ElevenLabs; AI actor generation uses fal.ai with ElevenLabs voiceovers, and social publishing uses Upload-Post. Those connected services receive requests for the features you enable.