More than 10 speakers to identify
Is there any plans to later support more than 10 profiles in the requests to identify speaker voices?
I am curious as to the technical limitations of the identification service. Is there plans in the future to make this a reliable service to make voice signature profiles out of?
Abe Millan
shared this idea
3 comments
-
We now support 50 speakers to identify in the identification API. Please kindly check it out.
-
Anonymous
commented
Also, don't just return a single profile id as the identification result. Please include at least 5 of them and their respective scores.
That way I can show them to the user which will then pick the correct one in case of ambiguity. I would then feed this info back into the model to improve its training.
The "High", "Normal" And "Low" confidence result set is just too coarse. -
Anonymous
commented
Yes, please, having just 10 profile is a serious limitation in my case. Suppose I want to detect which politician is speaking at the parliament. How can I do it with just 10 profiles?
