The voice you hear was artificially generated using speech synthesis software developed by ElevenLabs. The input was pure text and no corrections were made. Sign up: beta.elevenlab... Read our blog: blog.elevenlab...
Can’t wait for British accents, such as David Attenborough or Jeremy Irons. Imagine a book narrated by over 40 characters and all with depth, charisma and consistency of character, flawless immersion into the story, and loose yourself.
lol what are you talking about there is no depth or charisma possible with AI yet. We're far from that. It doesn't understand the context. What it can do, is convincingly simulate recognizable voices. It will not be able to match their charm, just their tics and mannerisms, which it will apply haphazardly to the text.
@Cameron L You're wrong if you think intonation is equivalent to charm. Again, the "AI" is just guessing at the most likely intonation for any given phrase. That's not charm, and these models are far from capable of the nuance, depth of character, or ingenuity of human artists.
This Is a terrible company. The voices sound awful. Their business practices are atrocious. Don't give them access to your bank account unless you don't want to have control over when and how much they charge you. I have cancelled my subscription twice after using it only once during the "free trial" and they are still trying to charge me like I still have a subscription months later. I called the number that was on my bill and it ended up going to somebody who said it was a wrong number. Then they called back and said that the company was started by his brother and he used his number to get the company as listed in New York, but he didn't know his number was going to be given out. I cancelled that same day and they have been trying to charge me still for the last 2 months. They don't allow any way for you to talk to a real person and everything is automated. You can 'clone' your voice. But it won't sound like you unless you are a robot. A cold-hearted robot, with bad business practices. DON'T TRUST ELEVENLABS!
I use Eleven Labs and love it so much. So impressive. I wish there was a tutorial section here for those of us wanting to use the API in various coding languages (c# for me). I'm quite aware that my c# skills are horrible and I know this is fairly simple to do but a 5 minute tutorial would literally take 5 minutes to make and post. Regardless - great work people!
Hi. Just wanted to say I really enjoyed playing with different AI voices. My favourite so far is Anthony. It's amazing to see how far technology has come.
Holy shit! I specifically chose The Great Gatsby to test out long passages with some of the voices I generated. Didn't think ElevenLabs would have made a video themselves to demonstrate their AI voice synthesis
Very Impressive, I can't believe the rate this sector is growing at. Best of luck for your future endeavours ! I have a feeling this will blow up like gpt did😅
The only way it properly blows up is if we're freely allowed to clone any voice we want. And it would blow up even bigger if we could map the generated voice to our own recorded (or live!) performance. Going a bit further, if we could do deepfake video with the replacement actor's voice, that would be great. Imagine Jim Carey's voice coming out of Jim Carey's face that's been faked onto Jack Nicholson's face in The Shining like we see on Ctrl Shift Face's channel. That video is still great with the original audio, but it would be even creepier with Jim's voice mapped onto Jack's performance.
@@mjt1517 Main reason it didn't blow up so far is because not many people are aware of ElevenLabs, which is actually good. And i don't think it should ever be completely free, for several reasons.
As an audio engineer with 30 years in the business, I have to say that this pretty darned impressive. For tge last couple of years I've been awed by Google interpreting my voice since I am Scottish and we have some considerably different sounds but Google, amazingly gets 80% or more correct. Since my job as an audio editor, listening to the tiniest nuances, I can attest to this incredible phenomenon that is AI. "Buccleuch" - was the only slightly off pronunciation made and it is a very Scottish name which we Scots pronounce as Bu-clue. It's truly a mind boggling thang !
Really impressive. The next step will be introducing finer control over the delivery, such as specifying when the text is meant to be read in a sad or angry tone, rather than the only option being to allow the AI to run wild with it.
The Eleven Labs site says they're working on that. Some day AI will not only read fiction convincingly, it'll generate background sound effects and music to enhance it!
cool. from hearing some other funny samples of characters from games and stuff, it certainly seems like a powerful tool! everyone will use it for different stuff I'm sure and there will be good and bad cases upcoming like there were with other voice ai programs. but hey, that's progress.
Just for fun, I had ChatGPT write a brief horror story in a first-person narrative and then had an Eleven Labs voice read it. It all sounded professional. I imagine I could add some music and sound effects and render something suitable for a commercial release!
It's use of intonation is phenomenal; trying to get my head around how well it understands "where and to what level" to use varying tones, its mind blowing!
This could end up being counted as royalties as they are literally still using someone's voice. But then laws and regulations regarding AI and Intellectual Property need to be expanded so while this is in gray area it could very well cause some chaos among VA's.
Indeed, it's the human performance and nuances that's pretty important. Maybe for more boring technical info readings AI will take over, but distinct narrative content will probably still be a human job
The voice generated sounds very real and human. My only criticism is that I've heard much better narrations of Gatsby, which isn't a fault of the technology itself but rather the voice chosen doesn't fit well.
Most of these comments are astroturf by employees. The voices are OK, but I'm sure they spend hundreds of thousands of characters to get just the right samples for this. I used this to make a few narrations 30 minutes to 50 minutes, and a 30 minute clip takes around 80,000 characters. I had to just accept so-so clips and move on because I didn't want to waste all of my quota regenerating the same text over and over again to get the perfect clip. Something like what they've done here would cost the average user around 300,000 to 400,000 characters. That would be 100 dollars per month to get a single 30 minute clip that sounds as good as this. They charge you for a fixed number of characters each month, but then delete your unused quota at the next billing cycle. They take WEEKS to respond to customer emails, and their emails have an attitude like "It doesn't matter what you think, other customers will replace you."
The android in Alien and Aliens never sounded robotic. And the androids in the movie AI sounded human. There are others. Nothing beats the HAL 9000, however.
This looks very good to me and I will get the $22 package to start! If it can do 50% of what it looks like it can do, it will be well worth it! Amazing!
Little heads up, since I've heard that voice before here in niche type of stories Channel by the name of Interstellar Roadtripper does use it for reading.
If Eleven Labs has an automated voiced AI it would be awesome. You wouldn't recognize it as a machine, except when it is laughing. Laughing probably is the bane of voice technology hahaha.
The speech is a bit harsh, but only in the way that quite a lot of narrators are (I'd probably prefer slightly softer speech - I'm British and sometimes struggle listening to American narrators so that's probably why). It is incredible, no doubt about it - you can tell that it's artificial but that doesn't matter, it does a great job of reading and making punctuation, sentence structure, and quotations clear. This is already good enough to be a commercial product from what I've listened to. Amazing work by everyone who's contributed to this.
I'm blind, and I already use it to have my own fiction read out loud to me, because having it read allowed helps me revise. It sounds 95% real. Also, they offer different accents, British included.
@@LaconicFlow Oh that's so incredible. I'm so glad that you get so much out of it. I imagine tech like this will really help with screen readers and things, but it's amazing that it's already useful in terms of accessibility stuff
For some reason, when I give it text to read, it starts out loud in every clip, but gradually gets quieter and quieter until it's almost whispering. This would be fine when you're only creating one clip, but when you're trying to chain several clips together for a long passage, the sudden, jarring jump in reading volume over and over, gets pretty ugly.
Yes, this sounds pretty nice. The issue with AI voices still is: While sounding like a human, almost every sentence follows the same "melody", and intonation which gets pretty boring quickly. For the time being: Without AI really "understanding" copy, it will be very hard for it, to give the copy "subtext", ie. the emphasis that makes the difference between "OMG! You look horrible!" (Subtext: "Are you sick?") and "OMG! You look horrible!" (Subtext: "You dress tasteless"). In order to create a gripping read, you need to understand the copy, its references etc. For the time being, I do not see how AI voices can fulfill this. It gets even more challenging if you want AI to understand direction, for example: "Yes, that was nice. Now add a little bit of red carpet. Sound celeb-style but still approachable."
That's where ChatGPT integration comes in. It will analyze the text and understand context, then parse that into tags/instructions that Elevenlabs will use to read correctly. It's already possible, they just have to create the interaction protocols.
@@danielrivera9147 Yes and no. So far, GPT does not really "understand" what you are talking about, it just spits out the copy that is most likely to follow a given prompt. And even if it does understand it. You will still need somebody, ie. a human being, that tells GPT what the changes should be. But if I need another human interaction, why not use a human in the first place that does it right in the first place …
@@chrispusnab And that true "understanding" will be extremely difficult to realize imho. Because real understanding means that you have to have the experience that teaches you the understanding. Difficult one for the time being, I think. I could be wrong. The speed of development is gobsmacking, but for now, the technology labeled "intelligence" is not that far.
You can clone your voice. That is one of the feature of a paid version. And there are others tools that allow for deep-fakes of your face with facial features.
And every other dead person with a few minutes of their voice recorded. You could literally hear the queen speaking original text now, if you uploaded samples to eleven labs website and did a voice clone. Brave new world.
Witam, dobrze było by gdyby tekst proponowany: 'First we thought the PC was a calculator. Then we found out how to turn numbers into letters and we thought it was a typewriter.' nie kosztował nic pod warunkiem że NIE jest edytowany
Prospective users of this should be cautious in reading the character limits. It's easy to misunderstand and think it means words limits. With character limits, every letter and number counts as a character, so "Come here" counts as 8 characters. They can be used up pretty fast.
I would really love to know your workflow for this. Did you just do it one paragraph at a time? How did you manage all the file names? Do you have an app to stick up all the paragraphs into chapters?
The diction is fluid and the prosody follows punctuation adequately. The voice timbre is realistic, as are the inflections. What's troubling is the overall tone - somewhat monotonous and too regularly-paced ; too insistent and lacking pathos.
I find some narrators can attempt to do this sort of monotone harsh speech, I imagine this is easier to synthesise but also what this is optimised for. It's amazing that the issues are so minor, it's great to live in the future
Depends on what the audio is trained on. If it's trained via common crawl it should be legal for commercial purposes given the contract binding it, but obviously the Getty suit against SD is testing this. If it's "ethically" sourced then it's entirely legal
Check out their tiers. You would probably want the Creator tier. $22/month for 100,000 plus 0.30 per 1000 additional characters. Or Independent Publisher if you were planning multiple novellas per month
Their site says they're coming out with a voice designer this month for users who want to create their own voices. Maybe you'll be able to create something you like.
I tried to reach out to you guys about one of your plans. I am wanting to use this to narrate an ebook and a course in the future. Not sure if you can only copy paste inputs and would the outputs be several audio files for an audio book?
@@blankblank2370 Ha, why don't you just go back to a cave somewhere and deny all the modern technologies that have made our lives so much easier? After all, it's so much better to use people to deliver messages and go to the library to seek information - it's certainly much more efficient!
@@blankblank2370 What if he's a small creator and doesn't have the funds to pay a voice actor? Or what if he just wants to do everything himself? None of us here owe voice actors anything. We're free to use their services or not.
Why do you not answer any of the comments? Have you seen how many people have asked you to share the voice character and settings you used! I guess this is why you only have 160 subscribers!...
