Join the waitlist
Early access · English version in development

The AI clipper that runs on your computer

No upload. No minute quota.

Turn long talking-head videos and livestream recordings into several short clips. Your video files stay on your machine, and local transcription and rendering aren't metered by the minute. AI clip planning reads only the transcript and is billed by its length, with an estimate before every run.

We'll email you when the English version opens. No spam, no sharing. Privacy

Planned for Windows 10/11 (64-bit) and Apple silicon Macs. Today the app is available in Chinese only.

Illustration. The current app is in Chinese; the English interface is in development.

Why local

Keep your footage. Skip the upload.

Many AI clippers run in the browser: you upload the whole file and spend minutes from a quota. This one does the heavy work on your own computer.

Your video files stay on your machine

Import, transcription (with the local engines), cutting and rendering all happen on your computer. Clip planning sends only the transcript text. The optional livestream product split sends still frames, never the video file.

No per-minute quota for local work

Local transcription and rendering aren't metered by the minute. AI analysis uses credits based on transcript length, and you see an estimate before you run it.

Built for long recordings

In the livestream workflow, multi-hour recordings are transcribed in segments. If it stops halfway, resume and only the remaining part is processed.

How it works

From one long video to several clips

This is how the current (Chinese) version works. The English version is planned around the same workflow.

  1. Import

    Drop in videos or whole folders — up to 100 files per project. MP4, MOV, MKV, WEBM, AVI, FLV, M4V, TS. Existing SRT subtitles work too.

  2. Transcribe

    On your computer with Whisper or SenseVoice, or with a cloud engine that's metered per minute and quoted before you start.

  3. Get clip plans

    AI reads the transcript and proposes 1–10 versions per run: highlights, condensed cuts, topic compilations or mixes.

  4. Review

    Play every segment, nudge in/out points, remove or reorder. Changes save automatically. Nothing exports until you say so.

  5. Batch export

    Export the versions you picked as MP4 — with captions, titles, music and intros if you want them. Up to 5 at a time, with hardware encoding when available.

For livestream sellers

Split a multi-product stream, then clip each product

A three-hour stream that sells ten products shouldn't take a day to cut.

  • Split by product first. AI looks at the video frames to find each product's segments and groups repeat appearances together. Uncertain boundaries are flagged for you to check. Metered by stream length, quoted up front, refunded in full if it fails.
  • Clip rules made for selling. The AI is instructed to use the most complete pitch for each product and not to stitch together prices or offers from different rounds. You still check before exporting.
  • Vertical by default. New livestream projects start at 9:16 with captions and a two-line title. Landscape recordings get a blurred fill. Change or turn off any of it.

Compared with web clippers

Where the work happens matters

Typical web-based clipperOxClipper
Your fileUploaded to the service firstStays on your computer
What you pay forUsually minutes of video processedLocal transcription and rendering aren't metered; AI analysis uses credits by transcript length
Long recordingsUpload time and minute capsNo upload wait; each AI analysis run has a transcript length cap (see below)
ControlAuto-generated clipsReview and adjust every segment before export

Straight talk

What it doesn't do

  • AI picks clips from the transcript, not from the picture. It doesn't judge faces, motion or "virality". (The optional livestream product split does look at still frames to recognize products.)
  • You review before you export. Some footage won't yield as many distinct versions as you asked for.
  • Each analysis run has a transcript length cap — for livestreams, roughly four to six hours of continuous speech. Longer recordings need to be split.
  • Clips don't tell speakers apart, so you can't ask for "only what the guest said". Transcription can mishear names, brands and numbers, so check them.
  • It's a desktop app for Windows 10/11 (64-bit) and Apple silicon Macs. Intel Macs and Linux aren't supported.
  • The English interface isn't released yet. That's what this waitlist is for.

FAQ

Questions

Can I download it today?
Not in English. The app exists and is in use in Chinese, but the English version isn't released. Join the waitlist and we'll email you when it opens.
Is it really local? What leaves my computer?
With the local transcription engines, your video and audio stay on your computer. To generate clip plans, the transcript text (with timestamps) is sent to an AI service. If you choose cloud transcription, the audio is uploaded to that transcription provider. If you use the livestream product split, still frames sampled from the recording are sent for product recognition. The video file itself is never uploaded.
Will it work with English-language videos?
The local Whisper engine transcribes English speech, and the transcription language can be set to English, Chinese, Japanese or Korean. The interface, presets and help are in Chinese today; bringing those to English is what the English version is for.
Is there a limit on video length?
There's no per-minute quota on local transcription and rendering. Each AI analysis run has a transcript length cap, roughly four to six hours of continuous speech for livestreams, so very long recordings need to be split.
How much will it cost?
English pricing hasn't been set. Waitlist members will hear about pricing, and any early-access offer if we run one, first.
Do I need a GPU?
No. It uses hardware encoding when your computer has it and falls back to the CPU otherwise. Speed depends on your hardware.
Which computers will it run on?
Windows 10 (1903 or later) and 11, 64-bit, and Macs with Apple silicon on macOS 14 or later.
What will you do with my email?
Use it to tell you when the English version opens. We don't sell or share it. You can ask us to delete it any time — see the privacy notice.

Be first to try the English version

One email when it opens. That's it.

No spam, no sharing. Privacy