The brief
Typing is the bottleneck, and speaking is about five times faster. We set out to build a dictation app that feels instant, works inside any app, and actually understands how each person talks, without sending your voice anywhere it shouldn't go.
What we built
- A real-time voice-to-text engine on Whisper large-v3-turbo that types into any app from a single hold-to-talk shortcut.
- An AI rewrite layer that strips filler, fixes punctuation and reshapes tone on a spoken instruction, with no copy-paste to a chatbot.
- A training system that learns your language: a personal dictionary for names and jargon, voice snippets for what you repeat, and custom autocorrections.
- 100+ languages with real-time translation, and audio that is processed live then discarded, never stored.
- The full app designed around it: dashboard, training, snippets and history, plus the on-screen dictation overlay.
The new interface: dashboard, voice snippets, history, and the dictation overlay states.
Stack
AI · Whisper
Electron
Product design
Design system
Windows + Mac
Next project · Websites
GolemType · the site