HOLD TO TALK.
RELEASE TO FINISH.
Hold the shortcut while speaking and release when you are done. NexVoice returns one polished paragraph to the current app.
VOICE WORKSPACE
Tap or hold to dictate. Double-tap to say what you want—without leaving your current app.
Hold the shortcut while speaking and release when you are done. NexVoice returns one polished paragraph to the current app.
Start with one press to begin recording. Speak at length to draft emails or documents, then press again to precisely transcribe your thoughts.
Select text on your screen or stay in your chat box. Double-tap to instruct it to summarize, translate, or draft a reply instantly.
the core questions
96.5%
of Chinese characters come out right.
We didn't pick perfect samples. We tested over 7,000 everyday recordings. This isn't lab data—it's the stability you'll feel every day.
96.9%
of English words come out right.
We applied the same strict standards to nearly 3,000 pure English recordings. This low error rate means you can confidently handle global meetings and emails.
0
recordings leave your Mac.
All speech recognition happens entirely on your device, with no cloud services quietly uploading your audio. Only when you explicitly ask AI to analyze or rewrite, the specific text is sent over the network.
before you begin
The practical details, without the fine print maze.
Tap once to start hands-free dictation and tap again to stop. Hold for push-to-talk and release to finish. Double-tap when you want to say an instruction such as “make this gentler,” “organize it,” or “translate it.”
Nowhere. Speech recognition runs entirely on your Mac, and there is no cloud speech service to fall back on—if the local engine cannot run, NexVoice stops instead of uploading your audio. Rewriting and answering are a separate step that sends only the text you asked it to work on.
No. NexVoice works alongside the input method you already use and writes the result into your current text field. Voice input should not force you to change how you type.
macOS 15 or later on an Apple silicon Mac. The recognition models run on the Apple silicon neural engine, so Intel Macs are not supported. Windows and mobile versions may follow, but there is no confirmed release date yet.
NexVoice is free for everyone in the current beta. Future pricing has not been decided; we will communicate any changes in advance.
Select “Apply for Beta,” enter your email, role, and industry, then submit the form. We review applications manually and send an invitation by email when a place is available.
start here
Download NexVoice and start with your next sentence.
macOS 15+ · APPLE SILICON · VERSION 0.2.98 · AUTOMATIC UPDATES