superwhisper is what you reach for when you want dictation that runs on your own machine and bends to your workflow. It transcribes locally by default, ships a mode system that rewrites your speech differently depending on what you're doing, and sells a lifetime licence instead of renting you the software forever. The trade is setup time: this is the most configurable on-device dictation app available, and it feels like it.
At a glance
Platforms
macOS 13.3+, Windows 10/11, iOS/iPadOS 18+
Pricing
Free; Pro $8.49/month, $84.99/year, or $249.99 lifetime; Enterprise on request
Free tier
Unlimited dictation on the small local models
Processing
On-device by default; cloud models optional via your own API keys
Languages
100+
Offline
Yes, fully
Integrations
Any text field; custom modes can trigger shell commands
Runs offline, properly. Local Whisper models mean audio never leaves the machine. This is the whole reason regulated professionals pick it, and unlike opt-in "privacy modes" elsewhere it's the default behaviour.
Modes. You define contexts — email, commit message, meeting note — each with its own prompt, model, and output format, and switch between them per keystroke. Nothing else in the category matches the configurability.
Bring your own model. Plug in your own OpenAI, Anthropic, or Google API keys and use whatever frontier model you prefer for the cleanup pass, rather than whatever the vendor picked.
Lifetime licence. $249.99 once. Steep against a $99 lifetime elsewhere, but it clears break-even against Pro monthly inside two and a half years.
Free tier is genuinely usable. Unlimited words on the small models, not a word quota.
Where it falls short
Setup is a project. The common complaint isn't a bug, it's that configuring superwhisper feels closer to standing up a server than installing an app. Budget an evening.
Proper nouns. The most consistently reported accuracy weakness. Custom vocabulary helps but you'll be feeding it names for a while.
Windows is much newer. The Mac app has years behind it; Windows shipped 1.0 in late 2025 and is visibly less polished.
No Android, no Linux. Android isn't on the roadmap.
Defaults aren't privacy defaults. Audio recordings are saved by default and API keys sit in plaintext on disk — fixable in settings, but worth knowing.
Pricing
Free gets you unlimited dictation on small local models — accurate enough for notes, not for a client email. Pro at $8.49 monthly or $84.99 annually unlocks the large local models, cloud models, custom API keys, file transcription, and translation. The $249.99 lifetime licence is the tier most people actually discuss. Students get 40% off and everything carries a 30-day refund.
Who it's for
People who dictate constantly, care that the audio stays on their machine, and are willing to spend an evening configuring something to get exactly the workflow they want. If you want to install an app and have it just work, Wispr Flow or Aqua Voice will be less painful. If you want the same local-first stance for a third of the price and don't need Windows, look at VoiceInk.
Last verified: 9 August 2026.
Similar to superwhisper
superwhisper: The power user's local dictation app, with a lifetime licence – Dictation Database
