superwhisper vs Wispr Flow.

The question that decides it

Start with where your voice goes, because it cascades into everything else. Wispr Flow’s privacy page states their answer: "Transcription always happens in the cloud to provide the best speed and accuracy." Everything Wispr does well follows from that sentence, including the four-platform consistency and the model sizes they can reach.

superwhisper built the other way round. Speech models run on your machine, a network is optional, and cloud models are a switch you flip when you want something larger. Their free tier exists because local inference costs them nothing per word, which is also why Wispr caps theirs.

For some readers that ends the comparison on the first page of the manual: confidential recordings and cloud transcription do not mix, and no feature list changes it. For everyone else it is one factor among several, in which case skip to the table.

Side by side

Wispr from their pricing and privacy pages and their help centre; superwhisper from their pricing page. All read on the date stamped above.

Where the money lands

Over three years at the annual rates, for one person, Wispr Flow costs about $432 and superwhisper about $255, or $249.99 once if you take their lifetime licence. Wispr prices per seat, so a team multiplies that side and not the other. That is not a small gap, and it is the clearest practical argument for superwhisper if the products otherwise tie for you.

The free tiers are further apart than the paid ones. On desktop, Wispr’s 2,000 words a week is a trial in everything but name, and one busy morning will finish it. Their Android tier is the exception, uncapped for now as a launch promotion. superwhisper’s has no word limit anywhere, which makes it a plan you can stay on indefinitely. Neither company is being dishonest about this; they are selling different things, and a cloud service has a per-word cost that a local model does not.

Pick Wispr Flow if

You dictate on Android, which superwhisper does not cover. You want one vendor running everything, with the same behaviour on every device and nothing to configure. Or your organisation needs the compliance paperwork: their ISO 27001 alongside SOC 2, and HIPAA available on all plans under a business associate agreement, is the strongest position of the two.

Pick superwhisper if

You want the option of working offline, or you would rather not send audio to a server by default. You want to transcribe recordings you already have, which Wispr’s app gives you no way to do. You want to choose the model, or bring your own API keys. Or you would rather pay once and stop thinking about it, which only superwhisper offers of the two.

The third option, and it is ours

Both of the above are cross-platform apps with AI rewriting at their centre. If what you actually want is narrower, there is a third shape, and this site sells it, so weigh the paragraph accordingly.

Edict is macOS on Apple Silicon only, $29 once, with the speech models inside the app and no cloud path in the product at all. It stays verbatim rather than rewriting, which matters when the text is a prompt an agent will act on, and it ships an MCP server, auto-send, and a wrong-window guard for exactly that job. File transcription with speaker labels is included rather than tiered. It has no phone app, no Windows build, no cloud sync, and fewer languages than either of them claims. If those absences read as missing features, one of the two apps above is your answer, and the full field has the rest.

Common questions