Topic
Measure of how correctly speech-to-text systems transcribe spoken language, typically quantified by word error rate and performance under noisy conditions.
A standalone AI recorder has to beat the free app on the phone it sticks to. 9to5Google's review of Comu's Action Pro shows that what the extra hardware buys you is where you are able to put it.
Reality
The scribe errors Healthwatch documented reached patients' records before anyone queried them. In two of the cases, it was the patient who did the querying. England has 27 products in use and no device classification.
Reality
A voicebot operator ran out of room stuffing client names into faster-whisper's initial_prompt, and now feeds it the system's own last 160 characters of speech instead.
Reality
The startup says a new model, Canto, will cut error rates from 30% to under 10% after weeks of user complaints. It is also entering meeting notes, where three incumbents already sit.
Perspective Coverage
3 publishersReality
The Series B values the dictation app at $2 billion and funds a push into meeting notes, a category where two of the four rivals Wispr names are already in California court.
Reality
Wispr raised $280 million at a $2 billion valuation six months after its last round. Its lead investor says the competitive set is every text box in front of an AI model.
Perspective Coverage
3 publishersReality