Privacy
Your videos, their audio, transcripts and everything you export are processed in your browser and are not sent to us or anyone else.
What stays on your device
- Every file you choose. The browser reads it from your disk; it is not uploaded.
- Audio extracted for transcription, transcripts, speaker labels, clip ranges, titles and caption edits.
- What the on-device AI reads and writes — the transcript text given to the language model and the titles it returns never leave the browser.
- Video frames processed by the background remover and the upscaler.
- Your exports, which go to your browser's downloads (and a temporary copy in browser storage, removed automatically).
- The result of the device check. It is shown to you and not stored or sent.
Every network request the app makes
| When | Destination | What is sent |
|---|---|---|
| Opening any page | freeclips.io | Ordinary page requests (HTML, scripts, styles). No media data. |
| When you start transcription, AI highlights, speaker labels or the AI upscaler | freeclips.io (/ort/) | Download of the ONNX Runtime engine (about 27 MB). We host it ourselves rather than loading it from a third-party CDN. |
| When you start background removal | freeclips.io (/mediapipe/, /models/mediapipe/) | The MediaPipe runtime and the selfie-segmentation model, hosted by us. |
| When you load the AI upscaler model | freeclips.io (/models/upscale/) | The Real-ESRGAN model you chose, converted and hosted by us. |
| First time you use speech, AI highlights or speaker labels (per model) | huggingface.co and its download CDN (*.hf.co) | Requests for the specific model files listed in our pinned manifest. Like any download, your IP address and browser details are visible to that host. No video, audio or transcript is sent. Any model file not in the manifest is refused. |
Every model file — ours or from Hugging Face — is checked against the SHA-256 in our manifest before it is stored or used; a file that doesn't match is discarded.
There are currently no analytics, advertising or tracking scripts. If we add advertising or analytics in future, they will be listed here, will not receive your media or transcripts, and the tools will keep working if you block them.
How we verify this: automated tests record every network request during page load, model downloads, transcription, AI highlights, background removal, upscaling and export with synthetic test media, and check that requests only go to the destinations above, carry no request bodies and never contain transcript text. That covers these flows in Chrome; it is not proof for every browser extension or environment.
What is stored in your browser
- Downloaded AI models, so you don't download them again.
- Temporary export files, removed automatically (24 hours after last use by default).
- Saved projects: transcripts, clips and settings for files you've worked on — never the media itself.
You can view and delete all of these on the storage page. Clearing this site's data in your browser settings removes them too.