Vosk seems to be where the brains/efforts behind the CMUSphinx project went, but I'm not sure -- it's been a long time (like ~10 years) since I used CMUSphinx but it worked reasonably well without much actual optimization way back when I used it -- it's possible/likely that the project is just mostly done and good at what it does. I think a lot of the team/people in the space have moved on to trying to use machine learning to solve the problem rather than the older model-based approaches.
Re:Vosk Speech Recognition Toolkit
by junon on Saturday May 22, 2021 @07:34AM
(#27245527)
This looks very approachable. Not in a position to test it at the moment - can anyone speak to its quality?
It shows that there are 50mib models for consumer devices or heavier models for servers - that's awesome and I've been looking for a project that packages things similarly.
https://news.ycombinator.com/item?id=27235970 (comment: https://news.ycombinator.com/item?id=27242148)
Vosk seems to be where the brains/efforts behind the CMUSphinx project went, but I'm not sure -- it's been a long time (like ~10 years) since I used CMUSphinx but it worked reasonably well without much actual optimization way back when I used it -- it's possible/likely that the project is just mostly done and good at what it does. I think a lot of the team/people in the space have moved on to trying to use machine learning to solve the problem rather than the older model-based approaches.
It shows that there are 50mib models for consumer devices or heavier models for servers - that's awesome and I've been looking for a project that packages things similarly.
https://ccoreilly.github.io/vosk-browser/