AI Video Tools
AI Video Tools
AI video editing that happens on your own device. A small model is downloaded into your browser once; the video itself never leaves it. 1 tool so far, free, with no account and no watermark.
AI video tools that run on your own device. Auto Captions transcribes speech with Whisper inside your browser and burns word-by-word animated captions into the video, with SRT and VTT alongside. The video never leaves your device; no upload, no account, no watermark.
Start here: Auto Captions (offline).
Make
How these differ from the AI for Business tools
The AI for Business tools send your text to a language model on a server, say so on every page, and count calls against a monthly allowance. These do not. The models here are small enough to run inside a browser — the speech recogniser is 41 MB — so they are served from this site, kept by your browser after the first visit, and run on your own processor through WebAssembly. The video is decoded, redrawn and re-encoded by the browser’s own media engine. No third-party server is contacted.
That is why there is no sign-in and no limit: there is no server bill to cover. It is also why the first run on a device takes a moment longer than the rest, and why a long clip takes about as long as it lasts.
What is coming to this section
- More languages for captions — the same model understands 99; they arrive once the English result has been proven.
- Silence and filler cuts — find the pauses and the ums from the transcript and cut them, with the words as the edit list.
- Clip from a long recording — pick a sentence in the transcript and export that stretch as a 9:16 clip with captions.
- Video compressor and GIF maker — re-encode on the device to a size a chat or a page will take.
The order depends on what people ask for. The contact page works, and so does the tool-request form in your account.
Frequently asked questions
Is anything uploaded?
No. Two downloads happen on first use — the model and the runtime, both from this site — and your browser keeps both. Your video is opened, transcribed, drawn and re-encoded on your device. We never receive it and could not look at it if we wanted to.
Why is the first run slow?
The model has to be downloaded once and the runtime warmed up. After that both come from your browser’s cache, and the time that remains is the work itself: speech recognition at about real time on a laptop, and re-encoding the frames.
Which browsers work?
Current Chrome, Edge, Safari and Firefox, on desktop and on phones. MP4 export uses on-device video encoding, which Firefox does not yet provide; there a clip is saved as WebM instead, with its sound.
Can I use the results commercially?
Yes. The output is yours. The speech model is OpenAI’s Whisper, published under the MIT licence, and nothing we make adds a watermark or a credit.