superwhisper vs Wispr Flow, and where Kana fits

These two answer the same question in opposite ways. superwhisper runs the speech model on your own Mac and will sell you a licence that never expires. Wispr Flow runs it on a server and sells you a seat every month. Almost everything else about them follows from that one decision.

The honest read on superwhisper

superwhisper is the pick if you want the recognition under your own roof and still want a say in it. You choose the Whisper model size, you can point it at a cloud model with your own API key when a local one is not accurate enough, and there is a lifetime licence at $249.99 so the bill eventually stops. It records and summarises meetings, and it runs on Windows and iOS as well as the Mac. What that flexibility costs you is an app with a lot of surface: models to download, modes to configure, and decisions to make before you have said a word.

The honest read on Wispr Flow

Wispr Flow is the most finished product in this category, and the writing is why. It reads the app you are in, infers tone without being told, and usually hands back text that needs no second pass. It covers 100+ languages, runs on Windows, iOS and Android, and it is the one of the two that has an answer for an IT department: a HIPAA-ready BAA on every plan, SOC 2 Type II for Enterprise, ISO 27001 under way. Your audio goes to a server for all of it, and the bill does not stop.

What actually decides it

Ask whether the audio is allowed to leave the machine. If it is, Wispr Flow writes better English than superwhisper does and there is not much left to argue about. If it is not, or if you work on planes and in rooms with bad wifi, superwhisper is the only one of the two still working.

Side by side

Where Kana fits

Kana takes superwhisper's side of this argument and then removes the choices. One model per language, one key to hold, a 6.5 MB app with no model picker and no modes. Dictation and file transcription are $9.99 once for three Macs, and AI rewriting is the only monthly part, because it is the only part that costs anything to run. What Kana has no answer for is Windows, meeting capture, a model you can swap, or the quality of Wispr Flow's first pass.

Which one to pick

superwhisper

You want local recognition and you want to tune it: model sizes, a cloud fallback on your own key, meeting capture, Windows and iOS in the same licence.

Wispr Flow

The output quality matters more to you than where the audio goes, you work across operating systems, or somebody at work needs the compliance paperwork.

Kana

You want superwhisper's local half with none of its settings, on an Apple Silicon Mac, for $9.99 once.

• superwhisper compared with Kana →
• Wispr Flow compared with Kana →