Open-source push-to-talk dictation for macOS. On-device, ultrafast. No account, no subscription, no cloud.
curl -fsSL https://digimata.github.io/parrot/install.sh | shpaste this in terminal, or download from GitHub
macOS 14+ · Apple Silicon
Audio goes straight to the Apple Neural Engine via CoreML, transcribes on-device, and types the result at your cursor. The first word lands in under 300ms — every utterance, every length.
Hold fn, speak, release.
Read the design notes →$ parrot
listening on fn hold · model: whisper-base.en
● recording
○ captured 2.40s · rms 0.041
→ 0.21s · the quick brown fox jumps over the lazy dog
● recording
○ captured 5.80s · rms 0.038
→ 0.27s · native swift, no electron, no python, no cloud round trip