Description

Noticed a lack of deep voices, this is mostly for realtime voice but can work with singing, trained on several minutes of normal speech, low notes (slight singing). Male friends tested and had to set pitch ranging from -8 to -14, will be lower than that for females (for the Your Reality cover, I had to do -21), so you'll have to play around with the pitch a bit. When using on kits.ai, I would recommend turning all enhancement settings off (compression, high pass etc.) compression makes my voice glitchy, and high pass cuts off parts of my voice due to it sometimes going below the threshold, all enhancement settings just don't work well with the model in general, and also setting conversion strength low since my articulation is weird and will not match up with most voices' pronunciation. Tested and made under Mangio-Crepe. < <