# rvc-support
- g
Gosmokeless28
05/25/2023, 10:52 AMYeah - j
josen95
05/25/2023, 10:56 AMThanks, good to know. So that means if I'm getting really bad results even though I have really good 6 mins of audio, it's prob something else. I used 200 epochs, is that a decent number? Or should I be cutting the audio clips to a certain length? I just used one long file if that's okay... - g
Gosmokeless28
05/25/2023, 10:57 AMFor 6 minutes of voice data, you should train an RVC model for 1,000 epochs, and each of the audio clips should not exceed 12 seconds in duration. - j
josen95
05/25/2023, 11:02 AMI see...okay noted on that. I should be cutting it up. Lastly, I'm thinking of increasing the quality by adding more clips to the dataset, but they'll be from other sources (so the mic quality and volume would be a bit different). Am I better of doing that or focusing on the 6 minutes from the same source that I have already? - g
Gosmokeless28
05/25/2023, 11:02 AMI'm sure it won't hurt the model - g
Gosmokeless28
05/25/2023, 11:05 AMNo problem - j
josen95
05/25/2023, 11:09 AMSorry, one more question haha. Would it hurt the model if some regular speech is included like in between the vocals? - g
Gosmokeless28
05/25/2023, 11:10 AMNope - t
Trevor
05/25/2023, 2:00 PMHi, I'm trying voice-to-voice but it is converting only 1-2 seconds. Is that normal? How can I render longer? - t
Trevor
05/25/2023, 2:01 PMI saw nowhere about a limitation what am I doing wrong? - g
Gosmokeless28
05/25/2023, 9:31 PMThat's odd - p
Prodabic
05/26/2023, 2:49 PMno matter what i do, i cant get the added file for the model im training, is there a way to fix it? - p
Prodabic
05/26/2023, 3:09 PMi just fixed it, i was doing it wrong - s
Scramble
05/27/2023, 3:16 AMI've trained two different datasets and for some reason it always NEVER sounds like the character - s
Scramble
05/27/2023, 3:16 AM - s
Scramble
05/27/2023, 3:16 AMexample of this - s
Scramble
05/27/2023, 3:16 AMthis for pepper, and with my second test of another dataset - h
huckleberrypie
05/27/2023, 9:40 AMIt depends on the audio to be inferred I guess - h
huckleberrypie
05/27/2023, 9:41 AMFor example I trained a Snow White SVC voice model with 200 epochs - h
huckleberrypie
05/27/2023, 9:41 AMHaving her sing "Virtual Insanity" yields only a smidge of her accent - h
huckleberrypie
05/27/2023, 9:42 AMbut if I feed in something that's just right for her vocal register like Halle Bailey's rendition of "Part of Your World" I do pick up Snow's voice - h
huckleberrypie
05/27/2023, 9:49 AMFor reference - h
huckleberrypie
05/27/2023, 9:51 AMThis one's from Adriana Caselotti but I do vaguely pick up the subtleties from what I can tell when I fed in Katie!Snow's voice - s
Scramble
05/27/2023, 3:51 PMIt can't be, because justinjohn proceeded to train pepper and he showed right and then used the same reference audio I did and it sounded better - s
Scramble
05/27/2023, 3:52 PM - d
Dingusmcbingus
05/28/2023, 9:58 AMHow good would an rvc model be with 25 minutes of data? - i
I Fucking Hate Papa Jons
05/28/2023, 5:28 PMive had models do well with 2 minutes