What it does
The project reports decisions in roughly 300 milliseconds per spoken word, allowing some actions to begin before the sentence ends.
Browser & computer use
jev-voice-browser
A voice-controlled browser where Jev selects an intent and target as speech arrives, then Playwright performs the action.
Bookmark: jev-voice-browser
Keep this in your collection.
Leave a noteWhat would you try with this?
: jev-voice-browser
Only you can see your notes.
Screenshot unavailable. Open the experiment ↗
How you can use it
You can adapt this approach to build hands-free web tools, such as an accessible browser assistant that clicks links or scrolls pages by voice. A developer can set up the software locally by providing an account key that connects the app to TypeSafe to understand spoken requests.
The system needs Chrome or Edge to turn speech into text. It also scans only up to one hundred visible items at once, meaning you must scroll down to reach lower links, and it cannot click content embedded from other websites.