Free local AI video clipper – No Upload, All in browser

A new free AI video clipper runs entirely in the browser, using local AI models to transcribe and clip long videos without uploading files. The tool leverages OpenAI Whisper for transcription and Qwen3.5-2B for clip selection, ensuring privacy by processing everything on the user's device.

Turn long videos into short clips, without uploading them Drop in a podcast episode, interview, lecture, or webinar recording and get back a handful of ready-to-post clips: each one a self-contained moment with a natural start and end, a suggested title, and optional burned-in captions. Everything runs in your browser. The video is never uploaded, there is no account, and there is no watermark. Cloud clipping tools charge monthly subscriptions and still make you upload multi-gigabyte files before the work can even begin. This tool skips the upload entirely, which for a two hour recording is often the slowest step of the whole process. How to clip a video - Drop a video file MP4, WebM, or MOV onto the tool. Talking content works best. - Click Find clips. The first run downloads the two AI models about 1.4 GB ; they are cached by your browser so every later video skips the wait. - Review the suggestions. Preview each clip, nudge its start and end points, toggle captions, then export the ones you like as MP4 files. What runs under the hood Three open source pieces power the pipeline, all executing locally. OpenAI Whisper MIT transcribes the audio with word-level timestamps, the same engine behind our audio transcription tool /transcription . Qwen3.5-2B Apache 2.0 , a small language model running on your GPU through WebGPU, reads the transcript and scores candidate moments for hook strength and completeness. Finally mediabunny, a WebCodecs toolkit, cuts the clips with hardware-accelerated encoding and paints the captions onto each frame during export. Because the language model is small enough to run in a browser, its picks are good drafts rather than a finished edit. That is why every clip comes with preview and trim controls: you stay the editor, the AI just does the reading. Private by architecture, not by policy Most video tools promise privacy in their terms of service. This one cannot see your video in the first place: there is no server doing the work. That matters for unreleased episodes, internal meetings, client interviews, course material, and anything under an embargo or NDA. It is also why it is free. We have no inference bill, so you get no meter. Frequently asked questions - Is this AI video clipper really free? - Yes, completely free and unlimited. The AI runs on your own computer inside your browser, so we have no per-video processing bill to pass on. There is no signup, no watermark, and no daily limit. - Does my video get uploaded? - No. The video file never leaves your device. Transcription, moment picking, and video cutting all happen locally in your browser using WebGPU. You can open your network tab and watch: after the one-time model download, nothing is sent anywhere. - How does it pick the clips? - First OpenAI Whisper an open source, MIT licensed speech model transcribes the audio with word-level timestamps. Then Qwen3.5-2B an open source, Apache 2.0 licensed language model reads the transcript and picks self-contained moments with a strong hook: a complete story, insight, tip, or exchange. You can adjust the start and end of every suggestion before exporting. - What video formats work? - MP4, WebM, and MOV files up to 4 GB. The exported clips are standard MP4 files. Talking content works best: podcasts, interviews, lectures, webinars, and screen recordings with narration. - How long can the video be? - A one to two hour podcast episode is the sweet spot. Transcription takes a few minutes for long files, and the clip picker reads the transcript in roughly ten minute windows, so even long recordings are covered end to end. - Why does it need WebGPU, and which browsers have it? - The clip-picking language model runs on your graphics card through WebGPU. Recent desktop versions of Chrome, Edge, Firefox, and Safari all support it. iPhones and iPads are not supported because iOS limits how much memory a browser tab can use, which is far less than the models need. - Can it add captions to the clips? - Yes. Every clip has a captions toggle. When it is on, word-timed captions are drawn directly onto the video frames during export, with the currently spoken word highlighted, the style short-form viewers expect. Because the captions come from the same Whisper transcript, the timing matches the speech. - Is this a free Opus Clip alternative? - It covers the core Opus Clip workflow, turning a long video into short clips with captions, with one big difference: cloud clippers make you upload the whole file to their servers and meter you with credits. Here the processing happens on your machine, so it is free, unlimited, and private. It does not yet do face-tracked vertical reframing. - Are the models open source? - Yes. Whisper Base is MIT licensed, Qwen3.5-2B is Apache 2.0, and the video cutting uses mediabunny, an open source WebCodecs toolkit. You could assemble the same pipeline yourself; we have just done the wiring, tuning, and UI for you.