Avette takes the best local ASR model and streams your voice directly into text fields — no pasteboard, no simulated keypresses, no Accessibility tricks.
Try it free on your Mac · €50, lifetime for Avette 1 · up to 15 devices
One recording — my actual voice — into macOS’s built-in dictation and into Avette. Hit play to hear it.
Avette doesn’t mess with your clipboard, doesn’t simulate keypresses, doesn’t use accessibility APIs to modify text inputs. It’s a proper input method: like a keyboard, except it doesn’t show up in the keyboard rotation (it’s actually a Palette IME!)
Pick your favorite microphone, whether your laptop is docked or on the go. Hide some inputs from the switcher if they’re never relevant. Switch from “push to talk” to “hands-free” with Right-Cmd. If you let go of the hotkey too soon, it waits for you to stop speaking. We could go on.
You’ll need an Apple Silicon mac (tested on M2 and later) because your audio is not being sent to any servers. Qwen3-ASR supports 30 languages and 22 Chinese dialects. Usable inside of any app: browser tabs, terminals, chat apps, rich text editors, etc.
Hi! I’m Amos — you might know me from articles/videos about Rust. I work too much, and regularly, my cat requires pets which prevents me from physically accessing the keyboard. There’s plenty of nice open-source dictation apps around (Hex, Handy etc.) but none quite worked for me: I wanted Qwen3-ASR, the best model right now in my opinion, I wanted to make it streaming (that’s genuinely novel tech I still hope to blog about), and I wanted to make it a real macOS input method, with a lot of care put into solving the little every day annoyances (input switching, dictation cut too early, etc.). I’m pretty happy with the result, and several people have asked me if they could have it — now the answer is yes!
— Amos
Secure checkout · VAT handled at purchase · invoice emailed instantly
Already bought? Manage your license →