Skip to content
ReviewRamble

The weekly brief

AI, reviewed, in your inbox.

  • New reviews of tools, models, MCP servers and repos, with the verdict up front
  • Re-checked prices and ratings when something changes
  • New analysis pieces as they are published

Models, agents, MCP servers and the tools actually worth your time.

One email a week. Unsubscribe with the link in any brief. How we use your email

AutoClip

A free, MIT-licensed OpusClip alternative that cuts long talks into captioned shorts on your own computer for pennies of model time.

Reviewed by ReviewRamble editors · Updated

View on GitHub Open source

Share this page:XLinkedInFacebookRedditWhatsAppEmail

Our rating

3.8/5

AutoClip does most of what a paid clipping service does for the price of a few model calls, provided you can live with a one-person project, Chinese-first docs and analytics that are on until you turn them off.

What is AutoClip?

AutoClip turns a long video, such as a podcast, interview or lecture, into short vertical or landscape clips with captions, a cover image, a title, a description and hashtags. You paste a link or a file, pick a platform (Douyin, Xiaohongshu, TikTok, Reels, YouTube Shorts, Bilibili or YouTube), and it renders up to 10 of the best-scoring clips that fit that platform's frame and length rules, keeping the rest as alternatives you can render on demand. Framing follows the speaker, and a ZIP publish kit comes with each clip.

Cutting, framing and rendering run locally. A language model of your choice reads the transcript to pick moments: a cloud API with your own key, or a local model through Ollama or LM Studio. Captions come from the creator's own subtitles, local Whisper or SenseVoice, or a cloud speech service. It ships as a desktop app for Apple silicon Macs and Windows, a Docker web app, and a CLI with an MCP server, all MIT-licensed and maintained by one developer.

Is AutoClip worth it?

AutoClip is the most complete open-source answer to OpusClip we have read about, and its economics are its argument. In the developer's measurements on an Apple silicon Mac, a 1-hour-43-minute English interview with existing subtitles became 10 Xiaohongshu clips in 7.5 minutes for about ¥0.09 of qwen-plus text-model time; a 2-hour-52-minute Chinese talk with no subtitles took 29.5 minutes and ¥0.20, 23 minutes of it local Whisper transcription. OpusClip's paid plans start at $15 a month for 150 minutes of uploads.

The pipeline is what a working editor would build. It scores moments from the transcript in a couple of model calls rather than the 128 sequential calls of its earlier version, keeps the speaker in frame and falls back to the full picture when nobody is in shot, and packages each clip with a cover, title, description and hashtags. The same pipeline is exposed through a CLI and an MCP server, so an agent can call `start_quick_output` on a folder of recordings. The project has about 12,800 GitHub stars and shipped v1.5.8 on 2 November 2026.

Privacy is better than a cloud service's but not absolute. Video never leaves your machine, yet cloud analysis sends captions and post text to your model provider, visual analysis sends sampled frames, and cloud transcription sends audio. Anonymous usage analytics go to PostHog in the United States and crash reports to Sentry, both on by default; the privacy policy is marked as a draft template awaiting legal review. API keys and platform cookies for publishing stay on your device.

The catch: it is one person's project. The README says so and warns there is no live support; five people have contributed code. Much of the documentation is in Chinese only, the Windows installer is unsigned, there is no Intel Mac build, and a brand outro is added to clips until you switch it off in Settings. The README also carries sponsored API-reseller links and an affiliate link to Alibaba Cloud. And clipping someone else's podcast from a link does not give you the right to post it.

Our take: for a creator or small studio cutting its own long-form recordings into shorts every week, AutoClip trades a subscription for an afternoon of setup and a model bill measured in cents; teams that need support, team seats or a signed installer should stay with a paid service.

Where it earns its place

  • A weekly interview podcast with a two-person team

    Turn each 90-minute episode into a week of Shorts and Reels

    Drop in the episode with its subtitle file, pick YouTube Shorts and Reels, review the 10 auto-rendered clips per platform and download the ZIP kits. Text-model cost is cents per episode against $15 a month or more for OpusClip.

  • A university lecturer publishing course highlights on Bilibili

    Cut landscape highlight clips from recorded lectures without uploading them to a third party

    Run the desktop app with a local model in Ollama and local Whisper so no audio or text leaves the laptop; turn off analytics and crash reports in Settings. The first local transcription install is about 220MB.

  • A developer building a content agent for a media company

    Let an AI agent process a backlog of recordings unattended

    Install with `uvx autoclip-mcp`, point the agent's MCP client at it, and call `start_quick_output` and `get_quick_output_status` per file; or script `autoclip produce talk.mp4 --srt talk.srt --platform douyin --json`.

Before you commit

Setup
Download the .dmg (Apple silicon) or Windows setup.exe, which bundle Python and FFmpeg; add a model provider key or a local Ollama or LM Studio model; a built-in sample project runs without a key. Linux and Intel Macs use Docker or the CLI (Python 3.10+ and FFmpeg).
Integrations
OpenAI-compatible APIs including Qwen, OpenAI, Gemini, DeepSeek, Doubao Seed, Kimi, GLM and Grok; Ollama and LM Studio; local Whisper, SenseVoice or Alibaba Bailian cloud transcription; direct publishing to Bilibili or through Upload-Post; MCP server and an agent skill.
Data and privacy
Video, audio and project files stay local; cloud analysis sends captions and post text, visual models get sampled frames, cloud transcription gets audio. Anonymous analytics (PostHog, US) and crash reports (Sentry) are on by default and can be switched off. API keys and platform cookies are stored only on your device.
Cost at scale
No licence fee. Per the developer's records, about ¥0.09 to ¥0.64 of text-model fees per long video depending on version and captions, plus cloud transcription at about ¥0.1 per audio hour if used; local rendering costs only machine time.
Limits
Interviews, podcasts, courses and talking heads are the validated inputs; low-dialogue footage needs a vision model. YouTube long-form requires clips of at least 180 seconds. Maintained by one person; no Intel Mac desktop build.

What’s good

  • Up to 10 platform-ready clips per video with captions, cover and post copy
  • Developer-measured text-model cost of ¥0.09 to ¥0.20 per long video
  • Cutting and rendering stay local; works fully offline with Ollama and local Whisper
  • Desktop app, Docker web app, CLI and MCP server share one pipeline
  • MIT licence, active releases

What’s not

  • Maintained by one person with no live support
  • PostHog analytics and Sentry crash reports on by default
  • Unsigned Windows installer, no Intel Mac build, docs largely in Chinese
  • Brand outro on by default; README carries sponsor and affiliate links

Best for

Podcasters, educators and small video studios who want to turn their own long recordings into shorts without a monthly clipping subscription.

AutoClip alternatives

AutoClip: common questions

Is AutoClip worth it?
AutoClip does most of what a paid clipping service does for the price of a few model calls, provided you can live with a one-person project, Chinese-first docs and analytics that are on until you turn them off. We rate it 3.8 out of 5.
How much does AutoClip cost?
Free and open source under MIT. You pay your chosen model provider: the developer's own runs cost ¥0.09 to ¥0.20 in text-model fees per long video, plus about ¥0.1 per hour of audio for cloud transcription.
What are the downsides of AutoClip?
Maintained by one person with no live support; postHog analytics and Sentry crash reports on by default; unsigned Windows installer, no Intel Mac build, docs largely in Chinese; brand outro on by default; README carries sponsor and affiliate links.
Who is AutoClip best for?
Podcasters, educators and small video studios who want to turn their own long recordings into shorts without a monthly clipping subscription.
How much does AutoClip cost for a team or in production?
No licence fee. Per the developer's records, about ¥0.09 to ¥0.64 of text-model fees per long video depending on version and captions, plus cloud transcription at about ¥0.1 per audio hour if used; local rendering costs only machine time.
Is AutoClip safe for company data?
Video, audio and project files stay local; cloud analysis sends captions and post text, visual models get sampled frames, cloud transcription gets audio. Anonymous analytics (PostHog, US) and crash reports (Sentry) are on by default and can be switched off. API keys and platform cookies are stored only on your device.
What are the best AutoClip alternatives?
The closest alternatives we have reviewed are VideoHighlighter, social-auto-upload and VideoAgent.

Answers checked Nov 21, 2026.