sVideo

A static video studio focused on privacy. Every operation runs in the browser, with no uploads, accounts, cookies, or analytics.

Features

• metadata removal by re-encoding the frames and stripping known GPS, EXIF, XMP, user data, and writing-app blocks from MP4, MOV, and WebM files;
• local detection of known ISO BMFF (mp4/mov) and EBML (webm) metadata blocks;
• pixelation, blur, and solid covers for selected areas, applied to every frame;
• resizing with optional aspect ratio preservation;
• bitrate control for smaller files;
• conversion between MP4 and WebM using the browser's MediaRecorder;
• anonymous filenames generated with crypto.getRandomValues;
• output verification before download (the clean copy is scanned again before it is offered);
• offline app shell caching with a Service Worker;
• responsive and accessible interface.

Run locally

Use a simple HTTP server because Service Workers do not run on file:// URLs:

python3 -m http.server 8080

Open http://localhost:8080.

Publish to GitHub Pages

This project includes .github/workflows/deploy-pages.yml.

1. Create a public GitHub repository named sVideo.
2. Push these files to the main branch.
3. Open Settings > Pages in GitHub.
4. Under Build and deployment, select GitHub Actions.
5. The workflow will publish the site automatically after every push to main.

With GitHub CLI:

git init
git add .
git commit -m "feat: create sVideo"
git branch -M main
gh repo create sVideo --public --source=. --remote=origin --push

If GitHub does not enable Pages automatically, select GitHub Actions in the repository Pages settings.

Privacy and limitations

The browser decodes the source file and exports a fresh copy through the Canvas API and MediaRecorder. The generated file is then scanned again and any remaining metadata blocks are stripped from the container before the download starts.

• Export runs in real time: a 1-minute clip takes about 1 minute to re-encode.
• The audio track is preserved by default. Audio plays during the export so it can be captured; keep the tab visible.
• Color, audio, and codec details may change slightly during re-encoding.
• Very large videos depend on browser memory and Canvas limits.
• The tool does not detect faces or text automatically. The user manually selects areas to hide.
• The debug page (?debug) exposes window.__sVideoTest with the scanner and strippers for manual verification.

Project structure

index.html                 interface and content
styles.css                 visual design and responsive layout
app.js                     local video processing
sw.js                      offline app shell cache
manifest.webmanifest       installable app metadata
favicon.svg                vector icon
.github/workflows/         GitHub Pages deployment

License

Licensed under the MIT License. See LICENSE.