arrivederci
  • Joined on 2023-04-28
arrivederci commented on issue ecker/ai-voice-cloning#152 2023-05-28 13:59:45 +00:00
VALL-E Integration (and In Response To TorToiSe: a Quick Retrospective)

Alright, sent you the link.

arrivederci commented on issue ecker/ai-voice-cloning#152 2023-05-27 21:26:35 +00:00
VALL-E Integration (and In Response To TorToiSe: a Quick Retrospective)

OK so I have a 2000 hour audiobook dataset compiled. Didn't take that long to gather but uploading it took forever. It's untranscribed still as well.

Use it if you feel like you're not making…

arrivederci opened issue ecker/ai-voice-cloning#248 2023-05-22 19:11:00 +00:00
Overfitting with large datasets
arrivederci commented on issue ecker/ai-voice-cloning#152 2023-05-22 14:20:43 +00:00
VALL-E Integration (and In Response To TorToiSe: a Quick Retrospective)

That actually looks encouraging. I'd give it some more time. Do you have a loss target in mind?

I do however wonder how it would fare if you gave it like 2000 hours worth of speech to train on…

arrivederci opened issue ecker/ai-voice-cloning#238 2023-05-15 22:24:04 +00:00
Another share your models thread.
arrivederci opened issue ecker/ai-voice-cloning#235 2023-05-11 09:45:15 +00:00
Graph not uodating after a day
arrivederci commented on issue ecker/ai-voice-cloning#152 2023-05-09 23:53:11 +00:00
VALL-E Integration (and In Response To TorToiSe: a Quick Retrospective)

Well I guess that's good news I guess, those metrics did look pretty bad

arrivederci commented on issue ecker/ai-voice-cloning#152 2023-05-07 12:23:14 +00:00
VALL-E Integration (and In Response To TorToiSe: a Quick Retrospective)
  • I think any noticeable jumps in the training metrics when I feed the beast will require an astronomical amount of new data, as I'm only at ~532 hours compared to the original paper saying it…
arrivederci commented on issue ecker/ai-voice-cloning#152 2023-05-06 20:04:20 +00:00
VALL-E Integration (and In Response To TorToiSe: a Quick Retrospective)

I was just playing around with vast.ai, a GPU peer sharing service and my first impression is that it works really well. Used it with the paperspace URL and it seems pretty robust.

You can get…

arrivederci closed issue ecker/ai-voice-cloning#221 2023-05-02 22:48:49 +00:00
Getting total gibberish when finetuning on a new language
arrivederci commented on issue ecker/ai-voice-cloning#221 2023-05-01 12:36:43 +00:00
Getting total gibberish when finetuning on a new language

OK nevermind it is actually producing pretty good output now, correctly pronouncing most of the words. I retrained on a 200 hour dataset of dutch audiobooks this night. The voice cloning doesn't…

arrivederci closed issue ecker/ai-voice-cloning#218 2023-04-30 12:14:48 +00:00
Memory leak in prepare_dataset() when using phonemizing using espeak (included temporary solution)
arrivederci reopened issue ecker/ai-voice-cloning#221 2023-04-30 12:14:17 +00:00
Getting total gibberish when finetuning on a new language
arrivederci closed issue ecker/ai-voice-cloning#221 2023-04-30 12:13:55 +00:00
Getting total gibberish when finetuning on a new language
arrivederci commented on issue ecker/ai-voice-cloning#221 2023-04-30 12:10:48 +00:00
Getting total gibberish when finetuning on a new language

4 epochs, I have added my train.yaml and tokenizer below. Currently also transcribing a 300h dataset to see if that helps.

Do you happen to know if the ipa tokenizer works for non english…

arrivederci opened issue ecker/ai-voice-cloning#221 2023-04-29 14:41:37 +00:00
Getting total gibberish when finetuning on a new language
arrivederci opened issue ecker/ai-voice-cloning#218 2023-04-28 11:13:02 +00:00
Memory leak in prepare_dataset() when using phonemizing using espeak (included temporary solution)