Why I built this: I saw WhisperFlow raised $80M and thought... wait, transcription is just running a model locally. Most Macs can handle this for free. So I built an alternative.
What it does: - Lives in your menu bar - ⌘T to start transcribing - See text stream live as you speak - ⌘T again to stop — auto-copied to clipboard - Paste anywhere
The tech that makes it possible: Instead of Whisper (chunked, slower), I used NVIDIA's Parakeet TDT. It actually streams token-by-token, not chunk-by-chunk. 6% word error rate vs Whisper's ~10%. Core ML optimized for Mac.
Replies
tldr.place
👋 Hey PH! I'm Tim, the maker behind YaprFlow.
Why I built this:
I saw WhisperFlow raised $80M and thought... wait, transcription is just running a model locally. Most Macs can handle this for free. So I built an alternative.
What it does:
- Lives in your menu bar
- ⌘T to start transcribing
- See text stream live as you speak
- ⌘T again to stop — auto-copied to clipboard
- Paste anywhere
The tech that makes it possible:
Instead of Whisper (chunked, slower), I used NVIDIA's Parakeet TDT. It actually streams token-by-token, not chunk-by-chunk. 6% word error rate vs Whisper's ~10%. Core ML optimized for Mac.