Web • AI
Scriptor
Scriptor turns speech into text without a server. Whisper, a speech recognition model, runs on the visitor's own graphics card through WebGPU, so recordings and videos never leave the computer.
The challenge was making a large model practical inside a browser tab. I used a compressed version of the model to shorten the download, kept it in the browser's cache so the next visit starts at once, and built a simple React interface around it: drop in a file, get the transcript.
Services
- Product design
- React
- WebGPU
- Whisper