The usual argument doesn't work here. Superwhisper also runs offline, also sends your voice to no server, and can also be paid for once. The difference is somewhere else.
01 · Common ground
Both apps think the same about the things that actually count, and that goes first, before looking for differences.
02 · The difference
Superwhisper uses Whisper, OpenAI's model, and to dictate you have to fetch one: between 60 MB and over a gigabyte depending on which, and you have to choose.
VoiceFlow uses the recognition engine macOS 26 already ships. That's why the download is 6.5 MB and it dictates from the first minute: what it needs came with the system.
The first time you open a dictation app you want to dictate, not compare models. That's the whole difference.
03 · And yet
It's in the settings, switched off. Anyone who wants it turns it on and downloads the same model they'd use in any other app: accuracy goes up and so does the weight, exactly the same.
We say so because it changes what this comparison means. It isn't «our engine against theirs». It's that with ours you don't have to fetch anything, and if one day you need to, it's there.
04 · Side by side
Theirs come from their site, with the date on it. Ours you can check by downloading the app.
| Superwhisper | VoiceFlow | |
|---|---|---|
| Works offline | Yes | Yes |
| Voice leaves the device | No | No |
| One-time purchase | Available | €59 |
| Ready to dictate after install | After picking and fetching a model | Right away |
| What it takes to start | Whichever model you pick | 6.5 MB |
| Languages | 100+ | 30 |
| Platforms | Mac, Windows, iPhone | Mac, plus iPhone companion |
Figures from superwhisper.com, checked on 26 August 2026. Whisper model sizes are the quantised ones from the whisper.cpp repository.
Try it
With 14 days of everything in Pro. No account, no email and no form: you download it and that's it.
Version 1.4.0 · 6.5 MB · Intel and Apple Silicon · requires macOS 26