Canto

What is Canto?

Canto is the first speech recognition model Wispr built itself, and it now handles dictation for every Wispr Flow user with nothing to switch on. Wispr started from a model pretrained on millions of hours of speech and text, then fine-tuned it, partly with reinforcement learning, to make fewer mistakes in hard conditions: noisy rooms, whispered or quiet speech, and one- or two-word phrases.

It is not a model you can call yourself: there is no API and no separate price, so using Canto means using Wispr Flow. Wispr has only published English results, and transcription runs in Wispr's cloud, not on your device. A successor more than ten times larger is in training.

Is Canto the same as Wispr Flow?

No. Wispr Flow is the dictation app; Canto is the speech model inside it that turns what you say into text.

The app

Wispr Flow adds the hotkey, dictation in any app, formatting, the dictionary, snippets, Notetaker and the plans you pay for.

The model

Canto does the speech recognition, and it is the first speech model Wispr built itself.

What it changed

Wispr says its own model cut error rates in the noisiest conditions from more than 30% of words to between 5% and 10%.

Switching

Nothing to turn on: every Flow user got it automatically, on every plan.

Who is Canto best for?

Canto is best for people who dictate with Wispr Flow away from a quiet desk, such as on a commute, in an open office or while speaking softly, and want fewer words to fix afterwards.

• Dictating in noisy places

  Wispr built and tested it for recordings with traffic, music, wind and other voices, where it beat the other real-time models it compared.

• Speaking quietly

  It tied for the lowest error rate on whispered and low-volume speech in Wispr's hardest test set.

• Names and jargon

  It reads the words in your Flow dictionary while transcribing, so names and rare terms come out right more often.

• Typing by voice in any app

  Through Wispr Flow it works in any app on Mac, Windows, iOS and Android.

What's included with Canto?

Tuning for noisy and quiet speech, dictionary hints for names, strong results on very short phrases, and data sharing you can switch off.

Tuned for hard audio

Reinforcement learning aimed at fewer errors with background voices, music, traffic, wind and far-off or whispered speech.

Dictionary hints

It takes the names and terms in your Flow dictionary as hints while it transcribes.

Short-phrase accuracy

It tied for the lowest error rate on one- and two-word dictations, where every model in Wispr's test made the most mistakes.

On by default

Every Flow user got it automatically, on every plan including Free.

Data sharing you control

Your audio and edits only help train Wispr's models if you allow it in Settings, and Enterprise accounts have training switched off.

How much does Canto cost?

Canto is not sold on its own: it runs inside Wispr Flow on every plan, from the free plan with 2,000 words a week on desktop to Pro at $15 per user per month for unlimited dictation.

Billed yearly, Pro is $12 and Growth $18 per user per month ($26 with unlimited Notetaker). Enterprise is priced on request, billed yearly, and adds SCIM provisioning, audit logs, MDM and model training switched off. Students and educators with an education email get 50% off Pro, and nonprofits get a discount. Canto reached every Flow user automatically, with nothing to turn on; Wispr lists no API or separate price for the model. Full pricing on wisprflow.ai. We checked these prices on 28 Sep 2026.

How do readers rate Canto?

No one has rated Canto yet. Be the first.

Canto pros and cons

Canto posts the lowest error rate in Wispr's own tests and copes well with whispers and noise, but it only works inside Wispr Flow, needs a connection and has only been measured in English.

What are the best Canto alternatives?

Willow Frontier Mini for unlimited free dictation, Superwhisper for offline use, and Gemini 3.5 Transcribe for an API you can build on.

Willow Frontier MiniFree

Choose it when you want free dictation with no weekly word cap; Willow's Basic plan gives unlimited use of it, though Willow itself calls it its weaker model.

SuperwhisperFree

Choose it when you need dictation that works offline; it runs Whisper models on your own device for free, and Pro costs $8.49 a month.

Gemini 3.5 TranscribeFree

Choose it when you are building your own app and need a speech API in 85+ languages, from about $0.005 a minute; Canto has no API, though it scored 3.4% to Gemini 3.5 Transcribe's 4.5% in Wispr's test.

Questions about Canto