Dataset to clone voices
#2
by
drewdru
- opened
Hi. I want to use ruslan.pth to clone other voices. I create dataset from audio podcast for 1.5 hours put it to voices/voice_name
Select ruslan.pth as Autoregressive Model and choose voice_name and click 'Generate', I got result too far from podcast.
How can I make this voice better? Should I run training? Or I can do it faster with voice cloning and bigger dataset?
Yes, training is necessary to get better results. I recommend having at least 1 hour worth of high quality void data. You can follow the steps outlined in https://git.ecker.tech/mrq/ai-voice-cloning/wiki/Training.
Thank you