Wispr Canto
Wispr's in-house speech model for messy real-world dictation — noisy rooms, whispers, and laptop mics.
About
Canto, released September 17, 2026 by Wispr's Advanced Interfaces Lab, is the company's first in-house speech model, built for messy real-world dictation: laptop mics, earbuds, traffic, music, whispered speech, and far-field recording. It replaces the previously undisclosed third-party ASR that powered Wispr Flow.
Wispr's release study used ten hours of opt-in English dictations from more than 2,300 speakers across apps and use cases, where the company reports the lowest word error rate among the models compared — though those internal recordings are not an independent benchmark. On a three-hour challenge set of noisy, low-volume, whispered, and far-field dictations, independent analysis found Canto ranked second overall behind Gemini 3.1 Pro and strongest among the real-time transcription models evaluated. Short one- or two-word utterances stayed difficult across all models.
Canto ships inside Wispr Flow, which dictates into any app in 100+ languages with a free tier of 2,000 words a week. Transcription is cloud-only even when users opt out of model improvement. Wispr raised $280M at a $2B valuation in August 2026.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Provider-run: lowest WER on 10h internal English dictation study; 2nd on 3h noisy challenge set behind Gemini 3.1 Pro.
Independent: no third-party rerun found.
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 7/10
Release facts (September 17, 2026; first in-house model; 10h/2,300-speaker study) confirmed by Wispr's research post and corroborated by independent analysis of the vendor's data. Docked for: (1) all evaluation figures are vendor-run with no third-party rerun; (2) no standalone API — Canto is only usable inside the Wispr Flow app; (3) transcription is cloud-only, which matters for confidential speech.
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 7.0/10
- Maintenance · recency, releases, issue load
- 8.0/10
- Openness · license + docs
- 2.0/10
- Manual review · sanity read + red-flag check
- 7.0/10
- · Adoption 7: Wispr Flow is a well-known dictation product ($280M raise at $2B, Aug 2026); Canto research post corroborated by independent analysis
- · Maintenance 8: first in-house model from Wispr's Advanced Interfaces Lab; successor >10x larger in training
- · Openness 2: proprietary; no standalone API — usable only inside Wispr Flow
- · Manual 7: 10h/2,300-speaker study design is public; all figures vendor-run with no third-party rerun