Meet Canto, our first speech model

Wispr Flow now runs on Canto, the company’s first in-house speech model. It’s rolled out for everyone, with nothing to switch on.

Wispr says Canto had the lowest word error rate among the models it tested on 9.8 hours of real-world English Flow recordings from more than 2,300 speakers. The comparison included models from Google, OpenAI, AssemblyAI and Deepgram.

On a separate test set with background conversation, music, traffic, wind and whispered speech, Wispr says Canto was the most accurate of the real-time models tested.