Open nto4 opened 11 months ago
@nto4 are you using random voices ?
@manmay-nakhashi hi, thanks for response, no ım usıng selected voice actually ım usıng fork of thıs repo for speed up https://github.com/manmay-nakhashi/tortoise-tts-fastest and ım usıng my audio sample for voice clonning
@nto4 I have seen sometimes it changes the voice , with some tuning in autoregressive temperature you can reduce that effect , I think.
thank you for answer @manmay-nakhashi can u provide me example because of ım very new in that field and just tryıng to learn
Im still facing this problem sometimes model decide use 2 speakers how can ı force use one speakers (im using this function and using my voice samples for voice clonning tts_withpreset i will be grate full if you can help me @manmay-nakhashi can u any advice ?
@n8bot you can consider it as one of the limitations of the model you can reduce it to some extent but it won't go away, fine-tuning further on cleaner data may reduce it further.
I am also facing the same problem. Can anyone help me? If I use a long text, there are points in the middle where a new voice is introduced
Hi @onlinerender, when using long text, the quality drops significantly. Please consider splitting up the text into chunks and generating those instead.
@fakerybakery can you suggest a cut off for number of words in a chunk?
hi, im trying to read some articles using tortoise tts but this model some times generating audio 2 different person speaking how can i prevent this ? i wanto to force tortoise read all text use same voice any body can help me ?
tts.tts_with_preset(parts[i], voice_samples=voice_samples, conditioning_latents=conditioning_latents, preset=preset)