The newly released checkpoint seems to have a significant gap in speech enhancement performance compared to the model described in the paper. This is also the case for the examples shown on the demo page.
Is this reasonable, or is there a problem with my testing?
Ryu1845
commented
1 month ago
The newly released checkpoint seems to have a significant gap in speech enhancement performance compared to the model described in the paper. This is also the case for the examples shown on the demo page.
Is this reasonable, or is there a problem with my testing?