IM modding Deus ex and had access to voice files so i added 25 files to 11 labs and it created the voice of JC Denton exactly how i wanted it. Took about two minutes. Very impressive stuff
For anyone wondering. He compared the two "basic" voice cloning features here. But the most important thing that should've been mentioned is Eleven Labs has this "proffessional voice cloning" which is unmatched. It uses up to 3 hours of recordings to train the model so it produces an obviously superior result.
Thanks for adding that important point. Yes, I compared instant voice cloning between the two. I wholeheartedly agree that elevenlabs professional voice clones are far superior to any instant voice clone.
Thank you for the video. I am a subscriber to ElevenLabs but find the UI in general a bit tricky. It seems to me that both Descript and ElevenLabs are more driven by IT technicians and programmer but are forgetting to make a nice, smooth and easy UI for users to choose and edit their voice options. It is tricky to find and twist the voices to what yiu really want. There are now companies building agents and web-sites where the choice of typical voiceover voices are much easier. ElevenLabs needs to wake up and make their services much more user friendly to use.
I am just trying Descript. Your comment is right on the money, it lacks flavor. It may sound like me but it doesn't feel, or taste like me. It doesn't know when I was excited about something or when I found something a bit humorous and wanted the end of a sentence to be lighter. But then how sincere do I sound on the 5th take?
Descript first had voice styles for their overdub feature, where you could give your sentences different feelings and intonations of your voice. But they somehow deprecated this feature, and I don't understand why. It's the main feature why I've paid 30 dollars each month to Descript for a subscription, because it was the only way to make my AI voice sound human; and they REMOVED it. I've tested to create a new AI speaker with my voice and I cannot create voice styles anymore and it sounds like "utter shit" now. Descript changed for the absolute worse.
Thank you. I just subsribed to ElevenLabs and created my voice. I found out that it sometimes pronounces the words wrong (in English). Not sure how to fix that. Do you have any idea why this happened and how to fix?
If you're using voice to voice, it should be pronouncing exactly as the submitted voice. If you're using text to voice, I've found elevenlabs to be very good with pronunciations, only occasionally missing some uncommon proper names. In that case, I spell the word phonetically and it gets it right.
I used 11 Labs for some time, but then they decided to just stop offering the Bella voice that my projects were using. That really drove home how you cannot rely in such a service. I was still planning on using them for cloning my own voice for YT vids (not for this channel) but now YT are saying such things have to be declared on every vid. That's... awkward. So nah, I'll just put text on the screen, or for shorter vids just record my own voice.
I've used the Creator version of Elevenlabs, which is pretty good. However, for longer scripts, it ALWAYS has errors somewhere in the final, and overall, AI isn't usable if you actually care about performance. They also overcharge for retakes as you try to fix their poor pronunciations. I used the Cloning exclusively and just canceled the Subscription because I did not want the VO to be obviously detectable. I would say it will be about two years before this type of product can really be used commercially and not by someone hacking at content.
Instant cloning requires 5 minutes of audio and is ready almost immediately. Professional cloning, which trains on a larger model and gets tested and refined, requires 30 minutes to 3 hrs of submitted audio and can take 4 weeks or more to be ready to use. With professional cloning, keep in mind that the submitted audio needs to be the same style of voice, environment, and recording conditions to get an awesome result. Hope that helps.
I would try the instant voice cloning and see what results you get. Meanwhile, you might want to go ahead and submit voice files for the professional cloning, since that will likely take a month or more for elevenlabs to train. You can work with the instant version while You're waiting for the pro version to get setup.
@@excelerator yeah that’s a good idea. One thing I’m not sure about cause I’ve only heard it from word of mouth. But apparently you will get flagged for uploading content with AI voices or AI content? Not sure if you know or not
I've used ai voices for videos on other channels here on UA-cam without any issues. In the examples I've heard of creators having problems that they assumed to be because of an AI voice, the real reason was something else - like copyright video clips or unoriginal content.
just to confirm, 11 labs is much better at creating custom AI voices than Descript? I think that's what you're conclusion was but it's a little confusing at the end - thank you for the video Subscribed and liked!
I've been doing these same tests, Have you tested the professional cloning form eleven labs yet? I'm still waiting on my creation to finish they say can take 4 week lol
I uploaded 3 hours of voice recording to ElevenLabs. In just a few hours it returned my professional voice. I was stunned at the result. My wife couldn’t tell it wasn’t me. Mind blowing. If you use 11Labs, you need to upgrade to get Pro Voice.
I just want to do voices for my Audiobooks I'm not sure if I would even use the other features. Will these voices pass in platforms like ACX because that is the goal. Thanks for sharing.
I definitely would NOT recommend descript for audiobooks. Elevenlabs, maybe. Test and explore to make certain you'll be able to generate a consistent voice for the length an audiobook would require.
Eleven labs would be better but isn't quite there yet. Give it a few months and it should be there. With that being said, there are TONS of bad real voice audio books already.
Won't pass -- ACX checks the file for most major AI voices like Eleven Labs and Descript and they generally get a result that the voice is +90% likely to be..... Also,, check the standard mp3 bits output -- they want 192 not 128 which most majors output unless you upgrade.
I can't recommend 11 labs for audio books, for 2 reasons. 1. They may drop the voice halfway through your book - they did that to me with the 'Bella' voice. 2. If you feed it too much text it gets very low volume and distorted towards the end. At best you can feed it one page at a time, and even then some voices will screw up towards the end of the page, so half a page at a time... The fact they swap and change the voices without any warning makes them impractical for bigger projects, and no, you won't get great results trying to clone their own AI voices. An AI voice of an AI voice sounds pretty bad.
Eleven Labs, Professional Voice Cloning is a process where you have to submit your files and wait 4 or so WEEKS. Its not a simple feature as they promote and sell it. Company didnt reply to my request for refund either. However they have an Affiliate program so suspect as often happens....
Thanks for sharing your experience. I did explain what I knew about professional cloning somewhere here in the comments. I'm sorry you didn't get a response from them - that's just crappy. I am an affiliate... But I cannot stand it when people hype up stuff for no other reason. I try to be fair and reasonable. And if I'm excited about something, it's because I'm really excited about it. For the most part, I try to keep things more of a show and tell - here's what I found and what it does - and you decide. I am curious to see how good the pro voice cloning is after 4 weeks... You would expect it to be pretty awesome, right?
I'm not sure about that. I have enough trouble with English - the only language I know. I'm in awe of folks who speak more than one language and especially those who learn English as a second language because it's so complicated!
I did a little digging. Marathi is not listed as one of their 29 supported languages for TTS (text to speech.) However, it is listed as an available language for dubbing (translating into another language.) I doubt that's what you're looking for. You might want to check out another really good platform, that does appear to support Marathi: excelerator.tech/GetPlayht
Sir thanks for the video .. I just wanted to ask if there is any software which allows you to change your voice so that : 1) . It sounds different from your original voice 2 ).It shouldn't sound robotic or chilidish or whatever changes done by adjusting the PITCH .
So I tried elevenlabs and its really good but the issue I have is it's not unlimited characters. I'm gonna hit the 100k characters in only a few days. Do you recommend anything else that might be unlimited?
I tried that but the voice clone wasn't as good as elevenlabs. I deleted it and am trying again. Waiting for voice clone to come back. Price is also a lot too unfortunately @@excelerator
Want to hear 4 voice cloning platforms compared? Check out this video: ua-cam.com/video/6ERJ-aQ3TdM/v-deo.html I compared Lovo AI, Speechify, PlayHT, and Elevenlabs. PlayHT was impressive!! PlayHT Unlimited Plan 70% off (Extended through 1/5/24.) 👉👉excelerator.tech/GetPlayht 👈👈 I'm not a fan of pushing "act now" promotions but, I noticed this discount this morning and it was NOT there earlier in the week, which leads me to believe it's a legit limited time sale and getting Unlimited for less than lower plan is a huge savings! I'm an affiliate, so if you make a purchase I may receive a commission. I sincerely appreciate you using my link.
ElevenLabs 👉 excelerator.tech/getelevenlabs
Descript 👉 excelerator.tech/getdescript
Elevenlabs is the clear winner in my opinion.
Same!
I agree. I use descript every day. But, when it comes to ai voices and voice cloning, they are nowhere near elevenlabs.
Nah my ai voice sounds more like me in descript. For example descript asks you to read specific words that I think make it better in training
@@Instant_Nerf Thanks for adding this perspective. Most people seem to have the opposite experience, but I'm glad it's working well for you.
@@excelerator that’s been my experience from the passed with both. It was a while ago. So maybe elevenlabs took the lead.
IM modding Deus ex and had access to voice files so i added 25 files to 11 labs and it created the voice of JC Denton exactly how i wanted it. Took about two minutes. Very impressive stuff
Glad to hear it did what you wanted!
For anyone wondering. He compared the two "basic" voice cloning features here. But the most important thing that should've been mentioned is Eleven Labs has this "proffessional voice cloning" which is unmatched. It uses up to 3 hours of recordings to train the model so it produces an obviously superior result.
Thanks for adding that important point. Yes, I compared instant voice cloning between the two. I wholeheartedly agree that elevenlabs professional voice clones are far superior to any instant voice clone.
Thank you for the video. I am a subscriber to ElevenLabs but find the UI in general a bit tricky. It seems to me that both Descript and ElevenLabs are more driven by IT technicians and programmer but are forgetting to make a nice, smooth and easy UI for users to choose and edit their voice options. It is tricky to find and twist the voices to what yiu really want. There are now companies building agents and web-sites where the choice of typical voiceover voices are much easier. ElevenLabs needs to wake up and make their services much more user friendly to use.
AI platforms, in general, haven't done the best at making cohesively connected and intuitive front-ends. Hopefully they improve!
I am just trying Descript. Your comment is right on the money, it lacks flavor. It may sound like me but it doesn't feel, or taste like me. It doesn't know when I was excited about something or when I found something a bit humorous and wanted the end of a sentence to be lighter. But then how sincere do I sound on the 5th take?
I can relate... On the 5th take I probably sound frustrated if anything!!
Descript first had voice styles for their overdub feature, where you could give your sentences different feelings and intonations of your voice. But they somehow deprecated this feature, and I don't understand why. It's the main feature why I've paid 30 dollars each month to Descript for a subscription, because it was the only way to make my AI voice sound human; and they REMOVED it.
I've tested to create a new AI speaker with my voice and I cannot create voice styles anymore and it sounds like "utter shit" now. Descript changed for the absolute worse.
Thank you. I just subsribed to ElevenLabs and created my voice. I found out that it sometimes pronounces the words wrong (in English). Not sure how to fix that. Do you have any idea why this happened and how to fix?
If you're using voice to voice, it should be pronouncing exactly as the submitted voice. If you're using text to voice, I've found elevenlabs to be very good with pronunciations, only occasionally missing some uncommon proper names. In that case, I spell the word phonetically and it gets it right.
@@excelerator it is text to speech. Good idea! Thank you.
I used 11 Labs for some time, but then they decided to just stop offering the Bella voice that my projects were using. That really drove home how you cannot rely in such a service. I was still planning on using them for cloning my own voice for YT vids (not for this channel) but now YT are saying such things have to be declared on every vid. That's... awkward. So nah, I'll just put text on the screen, or for shorter vids just record my own voice.
Wow, that had to be frustrating! I totally agree with you about becoming reliant for a critical component.
I've used the Creator version of Elevenlabs, which is pretty good. However, for longer scripts, it ALWAYS has errors somewhere in the final, and overall, AI isn't usable if you actually care about performance. They also overcharge for retakes as you try to fix their poor pronunciations. I used the Cloning exclusively and just canceled the Subscription because I did not want the VO to be obviously detectable. I would say it will be about two years before this type of product can really be used commercially and not by someone hacking at content.
so what do y ou use...i have long scripts myself
What is the difference between instant voice cloning and professional voice cloning?
Instant cloning requires 5 minutes of audio and is ready almost immediately. Professional cloning, which trains on a larger model and gets tested and refined, requires 30 minutes to 3 hrs of submitted audio and can take 4 weeks or more to be ready to use. With professional cloning, keep in mind that the submitted audio needs to be the same style of voice, environment, and recording conditions to get an awesome result. Hope that helps.
@@excelerator for just a voiceover on content. Which do you think would be best?
I would try the instant voice cloning and see what results you get. Meanwhile, you might want to go ahead and submit voice files for the professional cloning, since that will likely take a month or more for elevenlabs to train. You can work with the instant version while You're waiting for the pro version to get setup.
@@excelerator yeah that’s a good idea. One thing I’m not sure about cause I’ve only heard it from word of mouth. But apparently you will get flagged for uploading content with AI voices or AI content? Not sure if you know or not
I've used ai voices for videos on other channels here on UA-cam without any issues. In the examples I've heard of creators having problems that they assumed to be because of an AI voice, the real reason was something else - like copyright video clips or unoriginal content.
just to confirm, 11 labs is much better at creating custom AI voices than Descript? I think that's what you're conclusion was but it's a little confusing at the end - thank you for the video
Subscribed and liked!
You got it... Elevenlabs does a much better job. Sorry it wasn't clear but I appreciate the feedback.
I've been doing these same tests, Have you tested the professional cloning form eleven labs yet? I'm still waiting on my creation to finish they say can take 4 week lol
I have not. But I need to get some audio together and send it. You'll be way ahead of me in the line!! 😁
I uploaded 3 hours of voice recording to ElevenLabs. In just a few hours it returned my professional voice. I was stunned at the result. My wife couldn’t tell it wasn’t me. Mind blowing. If you use 11Labs, you need to upgrade to get Pro Voice.
Your AI voice sounded like an AI voice. I wished you had tried to produce the absolute best quality version of this.
You're right. I've recently learned quite a bit about what makes a good sample for cloning and I should do an update.
@@excelerator Yes you should.
Eleven Labs... hands down
Yeah, I'm kinda partial to ElevenLabs myself.
I just want to do voices for my Audiobooks I'm not sure if I would even use the other features. Will these voices pass in platforms like ACX because that is the goal. Thanks for sharing.
I definitely would NOT recommend descript for audiobooks. Elevenlabs, maybe. Test and explore to make certain you'll be able to generate a consistent voice for the length an audiobook would require.
Eleven labs would be better but isn't quite there yet. Give it a few months and it should be there. With that being said, there are TONS of bad real voice audio books already.
Won't pass -- ACX checks the file for most major AI voices like Eleven Labs and Descript and they generally get a result that the voice is +90% likely to be..... Also,, check the standard mp3 bits output -- they want 192 not 128 which most majors output unless you upgrade.
I can't recommend 11 labs for audio books, for 2 reasons. 1. They may drop the voice halfway through your book - they did that to me with the 'Bella' voice. 2. If you feed it too much text it gets very low volume and distorted towards the end. At best you can feed it one page at a time, and even then some voices will screw up towards the end of the page, so half a page at a time... The fact they swap and change the voices without any warning makes them impractical for bigger projects, and no, you won't get great results trying to clone their own AI voices. An AI voice of an AI voice sounds pretty bad.
Eleven Labs, Professional Voice Cloning is a process where you have to submit your files and wait 4 or so WEEKS. Its not a simple feature as they promote and sell it.
Company didnt reply to my request for refund either. However they have an Affiliate program so suspect as often happens....
Thanks for sharing your experience. I did explain what I knew about professional cloning somewhere here in the comments. I'm sorry you didn't get a response from them - that's just crappy. I am an affiliate... But I cannot stand it when people hype up stuff for no other reason. I try to be fair and reasonable. And if I'm excited about something, it's because I'm really excited about it. For the most part, I try to keep things more of a show and tell - here's what I found and what it does - and you decide. I am curious to see how good the pro voice cloning is after 4 weeks... You would expect it to be pretty awesome, right?
does is support Indian languge ( Marathi ) to?
I'm not sure about that. I have enough trouble with English - the only language I know. I'm in awe of folks who speak more than one language and especially those who learn English as a second language because it's so complicated!
I did a little digging. Marathi is not listed as one of their 29 supported languages for TTS (text to speech.) However, it is listed as an available language for dubbing (translating into another language.) I doubt that's what you're looking for. You might want to check out another really good platform, that does appear to support Marathi: excelerator.tech/GetPlayht
Thank you again
And, you're welcome again. Happy to see you coming back!
Sir thanks for the video .. I just wanted to ask if there is any software which allows you to change your voice so that :
1) . It sounds different from your original voice
2 ).It shouldn't sound robotic or chilidish or whatever changes done by adjusting the PITCH .
There is now! Elevenlabs recently released speech-to-speech and it does exactly what you're asking for ... www.excelerator.tech/elevenlabs
So I tried elevenlabs and its really good but the issue I have is it's not unlimited characters. I'm gonna hit the 100k characters in only a few days. Do you recommend anything else that might be unlimited?
Play HT is really good and their unlimited plan is on sale: excelerator.tech/GetPlayht
I tried that but the voice clone wasn't as good as elevenlabs. I deleted it and am trying again. Waiting for voice clone to come back. Price is also a lot too unfortunately
@@excelerator
Actually wait yeah why is it onsale for that link but when i go to their site normally it's not on sale@@excelerator
Well, I don't know 😁 but if the link gets ya the sale, that's good for both of us!
Decent. WhT. Was perfect
Descript sounds robotic.
Agreed! But, it's handy for fixing a word or two without having to re-record.
Want to hear 4 voice cloning platforms compared? Check out this video: ua-cam.com/video/6ERJ-aQ3TdM/v-deo.html
I compared Lovo AI, Speechify, PlayHT, and Elevenlabs. PlayHT was impressive!!
PlayHT Unlimited Plan 70% off (Extended through 1/5/24.) 👉👉excelerator.tech/GetPlayht 👈👈
I'm not a fan of pushing "act now" promotions but, I noticed this discount this morning and it was NOT there earlier in the week, which leads me to believe it's a legit limited time sale and getting Unlimited for less than lower plan is a huge savings!
I'm an affiliate, so if you make a purchase I may receive a commission.
I sincerely appreciate you using my link.
Eleven labs sucks at voice cloning
That hasn't been my experience.
@@excelerator voice cloning doesn’t exist, it is all hype, it is all synthetic