#VoiceLab - Professional Voice Cloning

1 messages · Page 1 of 1 (latest)

tawdry warren
#

Hi, I've been waiting a long while to try out the professional voice cloning on ElevenLabs, and just my luck I must have missed the processing by a day as had to wait just over 6 weeks to get a clone processed, this having to pay for two months subscription just to see if the platform is fit for the purpose I require.
The professional cloning was the only feature I was really interested in, but the results aren't as described.
The system describes the tool as "a perfect digital replica of your voice"
The "cloned" voice is not perfect, it's not even somewhat accurate.
I think this maybe due to the additional models used by eleven labs only being of upper class southern British people, as the result sounds like it could be me doing a very bad accent.
I provided over 25minutes of audio to the processor, and it hasn't given me the desired results.
The quick voice results in almost the same voice, sometimes its outputs a southern British accent, I've had it produce an almost Scottish accent and a couple of times its produced an Australian accent.
How can I get the system to produce a quality clone of my voice, I know it's possible but it just doesn't seem to be working for me, so I'm hoping it's something I am doing wrong and we can sort out?

drifting umbra
#

@tawdry warren Hi, thanks for reaching out. Firstly, sorry to hear you're facing this issue.

As per the official Elevenlabs documentation, users must make sure make sure you have enough material to clone the voice properly. The bare minimum we recommend is 30 minutes of audio, but for the optimal result and the most accurate clone, we recommend closer to 3 hours of audio. You might be able to get away with less, but at that point, we can’t vouch for the quality of the resulting clone.

#

So 25 minutes of audio is not enough to guarantee a quality professional voice clone.

#

"Provide at least 30 minutes of high-quality audio that follows the above guidelines for best results - preferably closer to 3 hours of audio. The more quality data you can feed into the AI, the better the voice clone will be. The number of samples is irrelevant; the total runtime is what matters. However, if you plan to upload multiple hours of audio, it is better to split it into multiple ~30-minute samples. This makes it easier to upload."

tawdry warren
#

I took another look and the audio provided was a 25minute clip cut down in smaller chunks, I think I had issues uploading and a larger file, so it was more than 25minutes, I'll have to see if I can get 3hours of audio then (thats going to be a long process of cutting up episodes, lol) and wait another 2 months 😦

drifting umbra
#

@tawdry warren I do believe it's possible for you to downgrade your subscription during the time your professional voice clone is being processed, and upgrade again when it's available to use.

#

That way, you save money

#

@azure pivot is this understanding correct? ^

tawdry warren
#

Hopefully the 3hrs will give me what I want then, I've got pretty good results with the same audio I uploaded to ElevenLabs from other services, but they provided a monotone, lifeless voice, and I've seen This service actually has better cadence and speech patterns, so was looking for something more true to life. So I'd assumed, wrongly, it would have been enough for this service too.
It was for a new project launching in January :S looks like it will have to wait 😢

azure pivot
#

Hi, @tawdry warren. Very, very sorry to hear about this. We are working on making the process much quicker (and even more accurate) in the the future, but it is a lot of research and development. Exactly what Voof said, the 25min is quite little data especially if you have more of a unique voice with a unique accent since it will require more data to clone that properly.

If you upload soon, you should not have to wait 2 months.

#

Yes, you should be able to downgrade and still have your voice cloned.

#

Could you provide some audio examples of the clone as well as the samples used for cloning? If you do not want to post them here here or in DMs, you can open a ticket.

tawdry warren
#

yeah I can send them over, this is the voice Id, I dont know if you have access to the voice in your admin system: vuhMk8xoBizCjZsVUzMq

#

I'll just locate the files I used for processing

azure pivot
#

Thank you very much! Listen to the first few sections, I can hear that there are a few issues. There are no pauses between the cuts which makes the voice's pace sound very unnatural. There are external noises that are not a part of your voice, for example, at around 0:42 and to 0:52. The other clips seem better, so should not be too much of an issue.

#

Unfortunately, I do not have access to anything in your account. Do I have your permission to try and create a clone from one of the samples you sent?

tawdry warren
#

Yeah, that audio is out in the wild anyway, I think its taken from the Tabs and Spaces podcast, so you can have a play with it.

#

Ah I see you're a mod for the discord and not an emplyee, I understand why you can't see any account stuff 🙂

azure pivot
#

I was going to ask about the podcast since I hear it is about games.

tawdry warren
#

its about coding 🙂

azure pivot
tawdry warren
azure pivot
#

From an outside view, that sounds very similar.

tawdry warren
#

its similar, but its still very southern and posh 😄 which is most def' not me 😄

#

I trialed a few recordings like this with a test group and they knew straight away it wasnt me 😄

azure pivot
#

Yeah, most likely much easier if one knows your voice like that. Then the only choice is PVC and uploading more data.

tawdry warren
#

yeah, I'll see what audio tracks are still on my H6 and see if I can pull off anything from there that can be used. I'll hopefully be able to get 3hrs or more from there

azure pivot
#

I really hope you can find enough data. Please keep me updated. Where is your accent from? Northern England?

tawdry warren
#

Im on the border of North Wales, so it's a bit of a weird one, its kinda nothern, kinda scouse, kinda welsh, and also non-descript all at the same time 😄

azure pivot
#

Haha yes, no wonder the AI is having trouble pinpointing your accent 😄

#

Please reach out to support, you can mention that J sent you.

#

They might be able to issue a complementary month or so since you are only waiting for this feature.

tawdry warren
#

thats even more upper class 😄

#

maybe its trying to tell me I need to be posh 😄 tbf I can sound like that, I'm pretty good at putting on an accent, maybe I should do that and then there wont be a problem lol

azure pivot
#

Maybe that is it :))

#

Sorry, I couldn't be of more help.

vast fossil
#

Please help, I'm unable to verify my own recording of my voice. The error says to get help from help center.

#

GSnVQwBlqJD2o1FdUN15

azure pivot
wicked bolt
#

Its really bad on 'English/British' accent It thinks everyone lives near London 🙂

tawdry warren
wicked bolt
#

Have you heard the 'Irish' one 🙂 LOL Its that comedy Irish Americans think everyone talks like there. It's painful.

wicked bolt
tawdry warren
#

Lol, American doing a bad Irish accent 🤣