This is mostly a curiosity thing, but when people contribute samples of them saying “Okay Nabu” to the Wake Word Collective, how/how often does the actual training translate to the actual performance of a wake word on, say, the Voice PE?
#Okay Nabu Wake Word Contributions
crimson stump
dreamy pond
This is mostly a curiosity thing, but when people contribute samples of them say...
To my knowledge (and I may be wrong...), The model is something that is worked on but releases are not done unless there is a significant improvements without drawbacks. Also I believe that it has been a while because a wall of sorts seems to have been hit in direct improvements. A more important improvement at this point would be able to filter a voice from background better.
There is work being done on voice but its a a complicated task.
crimson stump
Cool, thank you for the info!