Yehi AI is a local AI voice keyboard for macOS. Hold a shortcut key, speak, and Yehi types into any app. Your audio and transcribed text stay on your device, never on an external server. Free for personal use.
No reviews yetBe the first to leave a review for Yehi AI
Maker
📌
Hi Product Hunt,
Yehi AI is built for fast AI voice typing without sending private words to a cloud transcription API.
Yehi is a local AI voice keyboard for Mac. Hold a shortcut, speak naturally, release, and the text appears in the app already being used. It works across documents, email, chat, notes, code editors, and other places where typing happens.
What makes it different:
-Audio transcription runs locally on the device
-Voice and transcribed text are not sent to Yehi servers
-Text is typed directly into the active app
-Multiple local Whisper model tiers let users choose speed or maximum accuracy
-A local grammar model can clean up grammar, spelling, and punctuation after transcription
-Free for personal, non-commercial use
Yehi is especially useful for writers, developers, students, founders, journalists, and anyone who wants to write faster while keeping sensitive work private.
The current release is focused on macOS Apple Silicon. Windows is next, and the waitlist is open from the download page.
Feedback on whether the onboarding and model choices feel simple for a first-time user would be very helpful.
Report
Maker
One workflow where Yehi AI has become essential for me is working with Claude Code, Codex, Cursor and other text-heavy apps.
If you’re a developer, designer, founder, or builder, a lot of your day is spent writing prompts, notes, specs, emails, feedback, and rough ideas. Speaking those thoughts out loud is often much faster than typing them from scratch.
I’ve been using YehiAI as my own daily tool for the last 6 months, and this was one of the main reasons I kept building it.
Report
How well does it handle different accents or noisy environments, and does it work offline even without an internet connection at all?
Report
Maker
@hsan718483 It works offline, even with no internet connection at all.
For accents and noisy environments, the different local models help a lot. The larger models are better for tougher conditions like background noise, accents, or less-than-perfect audio.
Report
How well does it handle different accents and background noise in practice?
Report
Maker
@nuriyetoplp0p0 I use it while walking on the treadmill sometimes, and it works well. It still depends on the mic and noise level, but Yehi has different model choices, so you can pick faster or more accurate transcription depending on the situation.
Report
The hold-to-speak shortcut is such a natural fit for how I already dictate thoughts. Love that everything stays local too, no awkward server trust issues with personal notes.
Report
Maker
@sabrihakverdi Thanks, really appreciate that. I use it every day for my own work too, and the hold-to-speak flow has become the part I rely on most. Keeping everything local was important because so much of what we dictate is personal, unfinished, or work-related.
Report
the hold-to-talk shortcut feels really well thought out, especially how it drops the text right into whatever app you're already in without breaking flow. local-only processing for a voice keyboard is a nice move too
Report
Maker
@cengizmaskar Thanks Cengiz, really appreciate that. The goal was exactly that: hold, speak, release, and stay inside the app you were already using. The local-only part was important too because voice typing often includes private thoughts, work notes, and unfinished drafts.
One workflow where Yehi AI has become essential for me is working with Claude Code, Codex, Cursor and other text-heavy apps.
If you’re a developer, designer, founder, or builder, a lot of your day is spent writing prompts, notes, specs, emails, feedback, and rough ideas. Speaking those thoughts out loud is often much faster than typing them from scratch.
I’ve been using YehiAI as my own daily tool for the last 6 months, and this was one of the main reasons I kept building it.
How well does it handle different accents or noisy environments, and does it work offline even without an internet connection at all?
@hsan718483 It works offline, even with no internet connection at all.
For accents and noisy environments, the different local models help a lot. The larger models are better for tougher conditions like background noise, accents, or less-than-perfect audio.
How well does it handle different accents and background noise in practice?
@nuriyetoplp0p0 I use it while walking on the treadmill sometimes, and it works well. It still depends on the mic and noise level, but Yehi has different model choices, so you can pick faster or more accurate transcription depending on the situation.
The hold-to-speak shortcut is such a natural fit for how I already dictate thoughts. Love that everything stays local too, no awkward server trust issues with personal notes.
@sabrihakverdi Thanks, really appreciate that. I use it every day for my own work too, and the hold-to-speak flow has become the part I rely on most. Keeping everything local was important because so much of what we dictate is personal, unfinished, or work-related.
the hold-to-talk shortcut feels really well thought out, especially how it drops the text right into whatever app you're already in without breaking flow. local-only processing for a voice keyboard is a nice move too
@cengizmaskar Thanks Cengiz, really appreciate that. The goal was exactly that: hold, speak, release, and stay inside the app you were already using. The local-only part was important too because voice typing often includes private thoughts, work notes, and unfinished drafts.