(m)orpheus t(i)t(t)s

  • Uncensored Orpheus tts

finetune of orpheus on uncensored/ (un)alingned data to be able to generate more interresting sounds

SEASION2 - Experiment 3

speaker name is "baddy" - trained on base prob. final checkpoint for the time beeing

seems to even work with voice cloneing fine if you keep the speaker as baddy

bug reports / recommendations please in the discord https://discord.gg/RUs3uzBdW3

training still under way

does less tags but generalise rather well

Downloads last month
0
Safetensors
Model size
3.3B params
Tensor type
BF16
ยท
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support