metadata
license: mit
datasets:
- umoubuton/paimon
pipeline_tag: audio-to-audio
license: mit
datasets:
- umoubuton/paimon
abstruct
this is a so-vits 4.1 model of paimon(chinese)
"speech_encoder": "vec768l12"
more tainning paramater is in paimon_config.json
performance
this model need more trainning step(expected 40k)
it stopped at 20k now, with a demo song in folder demo, hope it can be trained further with stronger GPUs
thanks to umoubuton for the dataset