Does it support SSML?
#8
by
spaliwal
- opened
Can I control the audio output of the model with something like SSML tags? I see there's a support for non-verbals but they're only a subset of SSML tags. For example, how do make it say "1234" as digits vs a number?
The model is not able to generate a full range of SSML tags - we did add a few ones that should work relatively well. The fastest fix seems to be writing everything out. You can try "one thousand thirty four" vs "one two three four" and it will work.
But this is a valid problem, will think of ways to make it better