Spaces:

Fabrice-TIERCELIN

/

Text-to-Audio

like 11

Running

This is a very good work! Thanks so much.

1. Could we have the option to specify how many samples the user wants to generate instead of always generating a fixed number of samples per prompt?
2. Could we make the inference faster by using a different scheduler that does not need 100 steps for inference, and by using flash attention?

Thank you!