AutoClip: Let Qwen watch the boring parts so you don't have to

What it does AutoClip is a full-stack pipeline for turning long-form videos into short highlights. Feed it a YouTube or Bilibili link—or a local file—and it downloads the source, extracts subtitles, and runs a multi-step AI analysis to identify topic boundaries, score each segment, and generate titles. The output is a set of timestamped clips and suggested collections, served through a React web UI.

The interesting bit The analysis engine is built around Tongyi Qianwen via DashScope, not a generic vision model. It treats the video as a text outline first—extracting structure, then timelines, then scoring—before ever touching FFmpeg. That means the “intelligence” is essentially reading the subtitles like a script doctor rather than watching the pixels.

Key highlights

• Multi-platform ingestion: supports YouTube, Bilibili, and local uploads via yt-dlp.
• Async pipeline backed by Celery and Redis, pushing real-time progress over WebSocket.
• End-to-end workflow: outline extraction → timeline analysis → scoring → title generation → collection recommendation.
• React + TypeScript frontend with drag-and-drop collection editing.
• Dockerized deployment with a FastAPI backend and SQLite store.

Caveats

• Several headline features—Bilibili upload, subtitle editing, mobile support, and multi-account Bilibili management—are explicitly marked as in development.
• The AI analysis depends on an external DashScope API key and the Qwen model; there is no local or offline LLM option visible.

Verdict Content creators who regularly repackage long interviews or streams into short-form video will find the automated scaffolding useful. If you need pixel-perfect manual control or offline operation, this is not your tool yet.