Spaces:
Running
Running
live tts
#1
by
clem
HF staff
- opened
would be so cool to constantly generate as you write (maybe launch generation at every character & play at every space). Wonder if the voice can go faster than the writing
There might be a bit of variation across generations (since it uses a slightly different style vector depending on the length of the input sequence), but in principal this could work! Also, WebGPU is blazingly fast, so speed shouldn't be an issue. Might be an interesting avenue to explore for a community member! https://github.com/xenova/kokoro-web