I've done a coloured version, check it out:
https://mangadex.org/chapter/0002b028-f3f8-47c0-bba0-81860242a5a9
After page 26 (halfway), the quality improves as I did some changes that I thought would better work.
Had to split it into parts because the website doesn't take above 300MB.
So in ai speech generations, especially the local ones, there's always this annoying 'hiss' that I think is labeled as 'ai inference noise' or so. Is there a way to remove it? I tried using various ffmpeg ones, but the result is it works i guess, but it ends up making the audio sample sound really bad like through a phone, I need it to remain studio quality basically.
I then tried using UVR5 various AI models, but they don't even seem to pickup this 'noise' even with high agressive settings.
Anyone got any ideas?