Spaces:

akhaliq

/

VoiceFixer

like 52

Runtime error

For example this one here works great when they talk at the beggining, but the rest is destroyed completely, Idk if we could train another algorithm with good quality live performances, I guess you can't since you need a ground truth, there are some professional recorded concerts that are recorded by the audience as well, which are the only cases I could think of https://www.youtube.com/watch?v=pHuxFXmGWoY