Try it

The same two-stage pipeline the Mac app runs — speech to text, then a cleanup pass using the app's real prompts. Talk the way you actually talk: fillers, false starts, and all.

What's different from the Mac app: a browser can't load a 600 MB CoreML model, so the web version sends your clip to a hosted Whisper for stage one instead of running it on-device. The native app transcribes locally and only the cleanup step ever touches a network. Full breakdown →