autoclip

What it is

AutoClip is an AI-powered system for downloading videos from platforms like YouTube and Bilibili, automatically clipping highlights, and generating playlists. It uses a FastAPI backend, React frontend, Celery for tasks, Redis for caching, and SQLite (with guidance to upgrade to PostgreSQL) for storage. It includes a modular architecture with APIs for projects, clips, and collections, and supports real-time progress via WebSocket.

How it works

The system architecture links a user interface to a FastAPI backend, which delegates work to a Celery task queue, uses Redis for caching, and stores data in SQLite. The processing pipeline includes steps for outline extraction, timeline analysis, scoring, video generation, and collection recommendations. Frontend relies on React 18 + TypeScript; backend uses FastAPI, Celery, Redis, and SQLite; video downloading relies on yt-dlp; AI analysis relies on a model stack described as Tongyi Qianwen. The repository provides a mermaid diagram showing data flow: UI -> FastAPI backend -> Celery -> Redis -> SQLite, with F engine performing video processing, subtitles, and content understanding.

Getting started

• Docker deployment is recommended and requires Docker 20.10+ and Docker Compose 2.0+. Also needs minimum 4GB memory (8GB+ recommended) and at least 10GB storage. Typical one-click start commands are:

git clone https://github.com/zhouxiaoka/autoclip.git
cd autoclip

# Docker一键启动
./docker-start.sh

# 开发环境启动
./docker-start.sh dev

• Alternatively, local deployment uses start_autoclip.sh, quick_start.sh, status_autoclip.sh, stop_autoclip.sh. Manual installation steps include creating a Python venv, installing requirements, installing frontend dependencies, installing Redis, installing FFmpeg, and configuring a .env file. Key example commands:

# 1. 创建虚拟环境
python3 -m venv venv
source venv/bin/activate

# 2. 安装Python依赖
pip install -r requirements.txt

# 3. 安装前端依赖
cd frontend && npm install && cd ..

• The project exposes API docs at http://localhost:8000/docs and live frontend at http://localhost:3000, once running.

Recent releases

Latest release v1.2.0 (2026-06-03) introduces a one-click install macOS desktop client for AutoClip Desktop. It enables downloading videos, clipping, and local processing with no environment setup. Previous release v1.1.0 (2026-06-03) focuses on a fully installable macOS client with end-to-end validation and zero-dependency installation. The initial stable release v1.0.0 (2025-09-15) marks the project entering formal iteration, building on the autoclip_mvp feedback. The README shows these three release entries with descriptions in English and Chinese.

Traction

6282 stars, 1237 forks, 56 open issues as listed in the repo metadata. The latest activity shows last_push: 2026-06-03.

Behind the repo

The project is associated with a larger AI/Video processing stack leveraging yt-dlp for YouTube downloads, FFmpeg for media processing, and DashScope/Tongyi Qianwen for AI analysis, as listed in the tech stack and dependencies sections.

Caveats

License: MIT. Project started in 2025 and is actively updated as of 2026-06-03. Development notes mention features in development (Bilibili upload, subtitle editor) and the need to configure environment variables (DATABASE_URL, REDIS_URL, API keys) in .env. The configuration section provides example env variables and guidance for B站账号 management, which is under development.