Find your next open-source project

Search by name, purpose, or keyword — or paste a GitHub repository URL
Add a star history chart too →

zhouxiaoka/autoclip

Media & Design

Turns long podcasts, interviews and courses into platform-ready short clips with captions, covers and copy, locally rendered.

Best position on GitHub Trending (daily): #15. Click to get this badge for your README
11,052
1,928
Python
MIT
1 year old

What it does

AutoClip takes a link or local video and produces short-form cutdowns aimed at Douyin, Xiaohongshu, TikTok, Reels, YouTube Shorts, Bilibili and YouTube. It picks up to ten scored clips per run, adds subtitles, covers, titles, descriptions, hashtags and a ZIP publishing pack, and offers an editor when the automatic cut needs adjusting. The main targets are interview, podcast, course and talking-head footage.

The cutting, framing and rendering happen on your own machine. Analysis can use a cloud API, Ollama or LM Studio; subtitles can come from the original subtitle file, local Whisper/SenseVoice, or a configured cloud transcriber. Desktop, Docker/Web, CLI and MCP share the same quick-output pipeline. Sample outputs with source links are collected on the case library.

Quick start

The desktop route is the shortest. Download the build from Releases: .dmg for Apple Silicon macOS, -setup.exe for Windows 10/11 x64. The installer bundles Python and FFmpeg. On macOS 15+, if the first launch is blocked, allow it via System Settings → Privacy & Security → Open Anyway; on Windows, use SmartScreen's "More info → Run anyway".

Then:

  1. Open settings, choose a provider, enter your API key, pick an analysis model and test the connection. For local models, start Ollama or LM Studio first and select a model that service actually serves.
  2. With no project yet, click the built-in sample project to see the output without an API key.
  3. Import your own link or file, choose a portrait layout, preview the generated clips and download the publishing pack.

Intel Macs have no desktop package; use Docker or CLI instead. Docker deployment is cp env.example .env, create data, logs, uploads, then docker compose up -d --build, opening port 3000 for the web UI and 8000 for API docs. CLI/MCP needs Python 3.10+ (3.11 recommended) and FFmpeg/FFprobe on PATH; the release ZIP install path and the uvx autoclip-mcp one-liner are both documented in the quick start. A first CLI run looks like:

autoclip produce talk.mp4 --srt talk.srt --platform douyin --portrait-style podcast --json
autoclip outputs PROJECT_ID --export-kits

PROJECT_ID is the ID returned by the production step. Full setup details are in the installation guide.

Constraints

  • Desktop support is macOS Apple Silicon and Windows x64 only; Linux and Intel Mac users need Docker or CLI.
  • The software is MIT-licensed and free, but cloud analysis, transcription and AI cover generation bill against your own provider credentials. A local model avoids cloud API cost but needs suitable hardware.
  • Publishing directly requires a Bilibili or Upload-Post account.
  • Cloud analysis sends the relevant subtitles and copy; frame understanding or cover generation sends sampled frames; cloud transcription sends audio. Local analysis and transcription avoid those uploads. See the privacy note.
  • YouTube long-video mode requires complete segments of at least 180 seconds, so very short source material should target Shorts or Bilibili instead.
  • Local Whisper requires roughly a 220 MB first-time download.

HotRepo's take

AutoClip reached 15th on GitHub's all-language daily ranking on 2026-11-19, with 10,864 stars and 1,916 forks that day. Earlier daily records show 8,077 stars on 2026-09-21 (+266) and 8,617 on 2026-09-22 (+250), so the project had already built a substantial following before the top-15 day. That is real traction for a single-maintainer desktop tool in a crowded clipping niche, and the recorded gains suggest sustained rather than one-day attention. The README is unusually concrete about costs and runtimes: one 1h43m English source to Xiaohongshu took 7.5 minutes for ten clips with an estimated ¥0.09 in text-model spend, while a 2h52m Chinese source without subtitles took 29.5 minutes and about ¥0.20. Those are development-time measurements with different inputs and are not a controlled benchmark, but they set expectations better than most similar projects. The sponsored API resellers and affiliate link in the README are worth noting; they do not change the MIT code, but users should treat the provider claims as vendor marketing.

Written by AI from the project's README, updated . The project's own docs are the source of truth.

You might also like

  1. harry0703/MoneyPrinterTurboTurn a topic or keyword into a subtitled, narrated short video through an automated AI pipeline.Python142,911
  2. joeseesun/subcastDesktop app that turns overseas podcasts and YouTube videos into bilingual subtitles with AI summaries and Q&A.TypeScript104
  3. threerocks/article2mediaTurns an article or video into publish-ready Xiaohongshu cards, long images, WeChat layouts and narrated shorts via Claude Code or Codex.JavaScript126
  4. dreammis/social-auto-uploadCLI and Agent Skills for publishing videos, articles and image posts to multiple Chinese and global social platforms.Python16,539
  5. Bodila51/gtavi-claudecodeA Claude Code plugin for finding GTA 6 moments, cutting vertical clips and checking them against clipping campaigns.Python116