Short feedback
Hi there! Thank you for this incredible model, it really revolutionizes local music gen.
I did notice three quirks the model has. One, the sound is a little crunchy and a bit overdriven if you will, it seems the highs are too pronounced. Two, there's a clear bias for asian female singer voices, especially when you prompt it to generate electronic dance music of different kinds. So a bit more variety would be great here. Lastly, steerability could be improved for self-written songs, like it can miss quite a lot of words and even verses in the lyrics.
Once these three points are addressed, this is basically local Suno. On a side note, using reference audio also needs to be heavily improved as well.
Thank you again!
我是覺得可控性跟發音能更準更好，
我試做的我覺得聲音跟電子樂品質滿好的，
但現在好多需要有工具。