i have finetuned my model on roman urdu based on emotions at first i only passed emotion tags no ref codes which gave me good emotions but no cloning which is when i tried to finetune with refcodes as well which was a failure as well
Emotion only finetune gave me an output which worked on some voices but it leaked ref text to prompt text causing bad output
Emotion+refcode finetuning is giving me artifacts only with absolutely unrecognisible audio
I have tried emotion+refcode finetuning from scratch and also from the emotion only finetuned checkpoint
inshort i am facing problem cloning
help me figure out the issue and how to handle this
Thank you
i have finetuned my model on roman urdu based on emotions at first i only passed emotion tags no ref codes which gave me good emotions but no cloning which is when i tried to finetune with refcodes as well which was a failure as well
Emotion only finetune gave me an output which worked on some voices but it leaked ref text to prompt text causing bad output
Emotion+refcode finetuning is giving me artifacts only with absolutely unrecognisible audio
I have tried emotion+refcode finetuning from scratch and also from the emotion only finetuned checkpoint
inshort i am facing problem cloning
help me figure out the issue and how to handle this
Thank you