Changelog

What's new
in Tellaflow

Every update, in plain English. No 'performance improvements' without explanation.

v0.3.0Latest
March 2026
NVIDIA Parakeet support — parakeet-tdt-0.6b-v2 runs fully on-device
Live Translation overhauled — lower latency, more accurate on accented speech
Transcription speed improved ~30% on M2 and M3 chips
Microphone permission scoping tightened — only the active session sees mic access

We added Parakeet. NVIDIA's model is genuinely fast — especially for US English. If you've been running Whisper large and finding it slow, try switching. Also fixed the thing where Live Translation would occasionally emit the source-language word before catching itself. You know the one.

v0.2.0Major
January 2026
Live Translation shipped — speak any language, output English text
Whisper model selector added — choose from tiny, base, small, medium, large
Global hotkey support — trigger transcription without leaving your current app
No-internet mode lock — app now warns if a network request is attempted (it never should be)

The big one. Live Translation was the most-requested feature by a mile. Whisper's multilingual model is doing all the heavy lifting here — we just wired it to output the English translation instead of the source transcript. Works offline, works on planes, works in air-gapped offices. We also added model selection because 'large' is great but sometimes you want 'fast'.

v0.1.0Initial release
November 2025
Whisper (tiny, base, small) running fully on-device
Zero network calls — verified via proxyman
Global hotkey to start/stop recording
Transcribed text injected directly into the focused app

We shipped. The app does one thing: it takes what you say and puts it where you're typing. No cloud. No account. No subscription. We spent more time making sure it couldn't phone home than on the onboarding screen.

Ready to stop typing?

Free, forever. No account. No API key. Just your Mac and your voice.

Download Tellaflow free
Apple Silicon M1+ · macOS 13+ · 8 GB RAM