Whisper Large v3 Turbo
DefaultOpenAI
Covers around 99 languages and cleans up the filler as it goes, so transcripts read tidier than they were spoken. Pick this if you are not sure.
On-device. Offline. Private.
Wave Speak turns your voice into text, entirely on your Mac. Dictate anywhere you would normally type, or drop in a recording and get clean text back.
Free for 14 days. One-time purchase, no subscription.
Dictate anywhere
Hold your hotkey and talk. The moment you stop, your words are text, ready to drop into an email, a message, a doc, or an AI chat box. The whole loop takes about as long as the thought.
One sentence, two languages
Speak English and Chinese in the same breath and Wave Speak keeps up. The language is detected automatically, so there is no mode to set and no toggle to flip.
我review了一下你的design doc,整体approach没问题,但是scalability那一段要再补一下。
Strongest with Qwen3 for mixed English and Chinese. The default Whisper model covers 99 languages and switches between them too.
Turn recordings into text
Lectures, meetings, interviews, voice memos. Drop in audio or video and get clean text to read, or SRT and VTT subtitles with sentence or word-level timestamps.
Wave Speak Pro
Drop in a YouTube, podcast, or video link and Wave Speak Pro pulls the audio and transcribes it on your Mac, with the same models. No browser, no upload.
Pro only. The audio is fetched and transcribed on your Mac. If a video already has human-made captions, Wave Speak reuses those instead, even faster.
Three built-in models
Wave Speak ships with three of the latest open-source speech models, each from a different maker, all built in and ready from the very first launch. Pick the one that fits you.
OpenAI
Covers around 99 languages and cleans up the filler as it goes, so transcripts read tidier than they were spoken. Pick this if you are not sure.
Alibaba
Transcribes word for word and stays accurate when speech gets messy: noise, accents, more than one language. Especially strong on mixed English and Chinese.
NVIDIA
Near-instant dictation and transcription for English and 24 more European languages. Pick this if you speak one of them and want raw speed.
Add the names, brands, and jargon you use, so they come out spelled right.
Pick your own key combo. Hold to talk, or tap to toggle for long sessions.
A searchable record of your recent transcripts, ready to find and reuse.
Lives in the menu bar with no Dock icon. A minimal, out-of-the-way design.
Open source models from OpenAI, NVIDIA & Alibaba.
Private by design
Every model runs on your device. No account, no servers, no tracking. Your voice, your transcripts, and your files never leave the Mac. And because nothing depends on a connection, it all works the same offline: on a plane, in a dead zone, or with the Wi‑Fi off.
One-time purchase
Two ways to own it, both a single purchase with no subscription, and both starting with a 14-day free trial of everything.
Wave Speak
From the Mac App Store.
Apple Silicon Mac (M1 or newer). macOS 14 or later.
Wave Speak Pro
Direct from wavespeak.app.
Everything in Wave Speak, plus
Apple Silicon Mac (M1 or newer). macOS 14 or later.
Why are these three Pro-only?
Not by choice. Every App Store app agrees to the same safety rules, and Wave Speak follows them there. These three features need a little more freedom than those rules give, so they live in the direct version, downloaded from this site.
I built Wave Speak for myself first. I got tired of dictating in one app and copying the result into the one I actually needed, and I did not want a meter running every time I used it. So I made the version I wanted: that same quality of dictation, landing right where you are already typing, running entirely on your own Mac. You can write as fast as you think. I hope it takes a little friction out of your day too.
The maker of Wave Speak