Автонарезка видео на вирусные Shorts и Reels

autoshorts is an open-source desktop application that automatically cuts long video or audio recordings into high-impact vertical short-form clips (9:16 portrait) with AI-powered viral moment ranking. Built on Tauri 2 + React + Rust + SQLite, it keeps all data — transcripts, clip candidates, render metadata — stored locally on the machine. Reach for it when you have hours of podcast, stream, or interview footage that needs turning into Shorts or Reels without manually scrubbing through the timeline, or when uploading raw footage to a cloud service is not an option. The pipeline runs automatically: import media → extract audio → transcribe via Deepgram or local Whisper → rank moments with DeepSeek, Claude, or local Ollama models (LLaMA 3.2, Qwen 2.5 3B/7B) → auto-crop landscape to portrait H.264 using ffmpeg. A native project manager handles multiple projects from the dashboard. Direction is clip extraction, not timeline editing or social publishing — for scheduling and posting, pair it with a separate tool. Prebuilt installers cover macOS Apple Silicon and Intel (.dmg), Windows (.msi/.exe), and Linux (.deb/.AppImage). Prerequisites: ffmpeg in system PATH; offline transcription additionally requires Python and openai-whisper. Cloud mode needs API keys for Deepgram, DeepSeek, or Anthropic.