AutoClip
GitHub9k
Download
Open source · Free · Local

One link.
One click.

Paste a link, pick platforms, press once. Get clips, captions, covers and post copy together. Edit and render on your computer.

Download for macOS See real outputs
Also for Windows · Docker / CLI / MCP · v1.5.0 · Open source, MIT · Free

That one link →

One click

No editor.
No chatting with an AI.

Clipping used to mean one of two things: dragging a timeline frame by frame, or explaining what you want to an AI, again and again. AutoClip removes both.

Video editor

Find the moment, drag the timeline, fix captions, style them, export. Then do it again for the next platform.

AI chat / agent

Explain what you want, answer its questions, correct it, explain again. There goes the afternoon.

AutoClip
  • Paste a link, pick platforms
  • Press once
  • Get clips, covers, titles, descriptions and tags

Editing and rendering stay local. Cloud services are billed by your selected provider.

Case library

One link. 10+ clips ready to post.

Real, untouched AutoClip outputs, updated as we go. Open any clip to copy the post text and save the cover and video — ready to publish. Made something good? Share it.

How it works

You do the first step.

  1. 01Paste a link, pick platforms
  2. 02It reads the whole thing and picks
  3. 03Packaged per platform
  4. 04A full publish kit

Paste a link, pick platforms

A YouTube or Bilibili link, or a local file. Pick any of Douyin, Xiaohongshu, Bilibili, TikTok, Reels, Shorts and YouTube, and add more later without starting over.

https://www.bilibili.com/video/BV1B5xkzPEhx
DouyinXiaohongshuBilibiliTikTokReelsShortsYouTube

It reads the whole thing and picks

The full transcript is read in one pass to find complete questions and answers: starting at the first line of the question, ending when the answer is done and the speaker pauses. Creator subtitles are used directly; otherwise it transcribes locally.

    A 2–3 hour interview usually yields 18–33 moments · the top 10 are made automatically · the rest wait for one click

    Packaged per platform

    Framing follows the speaker, foreign speech is translated to the platform’s language, and colour and caption motion follow each clip’s mood. Douyin and TikTok each get their own look.

    InterviewDouyin · Xiaohongshu
    PodcastTikTok · Reels · Shorts
    The model reads each clip’s mood (calm, serious, bold, sincere or light) and picks from 7 palettes and several caption motions. Clips vary; each clip stays consistent across platforms.

    A full publish kit

    Every clip comes with a title, description and tags written for the platform’s rules, plus a matching cover. Copy the post text or download the kit, and publish.

    Publish kit · Douyin
    clip.mp4cover.jpgpost.txt
    Copy post textDownload kit
    Speed & cost

    Faster, and cheaper.

    This version picks clips in one pass instead of over a hundred model calls, runs calls in parallel, uses creator subtitles when they exist, and renders with your computer’s hardware encoder. Measured on the same Mac.

    Why AutoClip

    Local editing, your choice of model.

    AutoClipSubscription cloud tools
    CostFree software; cloud services billed by usageSubscription or usage pricing; check the chosen service
    Where your video goesEdit and render locally; selected cloud features receive subtitles, sampled frames or audioProcessing location and upload scope depend on the service
    ModelsYour choice: Qwen, DeepSeek, GPT, Gemini and more, or local OllamaAvailable models and settings depend on the service
    Chinese platformsTemplates for Douyin, Xiaohongshu and Bilibili, with Chinese captions for foreign sourcesCheck platform templates and language support
    Code & automationMIT open source, hackable and self-hostable; CLI / MCP for agentsCode availability, API access and plans vary by service

    The right column summarises public pages of comparable cloud products as of October 2026. Check each product’s latest terms.

    Developers

    One command to clips, or hand it to an agent.

    The desktop app, CLI and MCP server share one pipeline and one data folder. Projects made from the command line show up in the desktop app.

    • CLIBatch jobs, with --json output for scripts and pipelines.
    • MCPSeven tools that Claude, Cursor, opencode and other agents can call to make clips.
    • DockerSelf-host the web app on a server or NAS.
    # Use your saved model settings, one command to clips $ autoclip produce talk.mp4 --platform youtube_shorts --json # Inspect outputs and export publish kits $ autoclip outputs PROJECT_ID --export-kits # Connect opencode or another MCP client $ autoclip mcp install opencode

    CLI / MCP docs →

    FAQ

    The questions you were about to ask.

    Reproducible bugs: GitHub Issues. Discussions: Announcements, first clip, ideas and roadmap. Known issues: pinned issue.

    Stuck on your first clip? See the first-clip Q&A.

    Does it cost anything?

    AutoClip is free under the MIT license. Cloud analysis, transcription and image generation are billed by your selected provider. On 2026-10-01, a 1h43m Jensen interview with creator subtitles produced 10 clips on an Apple Silicon Mac using qwen-plus, with an estimated ¥0.09 in text-model fees. Cloud transcription, image generation and publishing-service fees are excluded. Local Ollama / LM Studio analysis has no cloud-model fee and needs suitable hardware.

    Is my video uploaded anywhere?

    Editing, framing and rendering run on your computer. Cloud analysis receives relevant subtitles and copy; optional vision or reference-frame image generation receives sampled frames, and cloud transcription receives audio. Local analysis and transcription do not require the corresponding cloud APIs. Finished clips upload only when you choose to publish to a connected platform.

    Which sources are supported?

    YouTube and Bilibili links, and local files. Use creator subtitles when available; otherwise use local Whisper / SenseVoice or configured cloud transcription.

    Why does the OS warn me on first launch?

    The builds are not notarized (macOS) or code-signed (Windows) yet. macOS: right-click the app and choose Open. Windows: More info, then Run anyway. It only happens once.

    What if the picks aren’t good?

    Try a stronger model first. If results are still off, tell us the video type, length and model in the pinned GitHub issue. We collect samples to improve clip picking.

    How should I try it the first time?

    Start with a 10–30 minute interview or podcast that has subtitles, and pick one or two platforms. If you get stuck, see the first-clip Q&A.

    How do I publish to Bilibili and global platforms?

    Add an Upload-Post key and a Bilibili cookie in Settings, pick platforms on a clip, and keep Only me for a first try. Step-by-step: publish guide.

    Download

    Your next great clip starts here.

    macOS and Windows. Free. Python and FFmpeg are bundled, so there is nothing else to install.

    macOS

    v1.5.0 · 270 MB

    Apple Silicon (M-series) · macOS 13 or later. Intel Macs are not supported yet.

    Download .dmg
    First launch

    The build is not notarized by Apple yet. On first launch, right-click the app and choose Open. It opens normally after that.

    Windows

    v1.5.0 · 196 MB

    Windows 10 / 11, x64. Per-user install, no admin rights; the installer fetches WebView2 if it is missing.

    Download installer
    First run

    The installer is not code-signed yet, so SmartScreen warns once. Choose More info, then Run anyway.

    Running on a server or NAS? Docker and source setup are in the GitHub README.