Warning: Using Ai Voices On Youtube Is Very Dangerous

OMG Thats awful! All your work! I use youtube only for watching the videos, but was thinking about it the other day. Thanks for the info Brian.

Funny thing - I was reading an article just the other day and it said NO ONE likes the sound of their own voice. They gave reasons for it, of which I immediately forgot, but leads me to believe your own voice is just fine.

It looks like YouTube is cracking down on AI content. https://blog.youtube/inside-youtube/our-approach-to-responsible-ai-innovation/

A quick Google search revealed a lot of people discussing YouTube crackdowns on AI voices, too.

Now that you know what the issue is, are you going to redo the videos without AI voices?

I wonder what signature an AI voice has that the system can recognize it and why of all things that is against the rules.

I used copyright free music, of an old song, something classical. It matched it with something a normal band played, but the song itself was well out of copyright. But annoying.

Personally I think its dangerous to use your own voice because the voice can be stolen and used in scams. And it doesn't take much. A few seconds. I don't care for my voice, some people love the sound of theirs, can't imagine why.


I wonder if there is an AI voice maker you can use, like train it on your voice but then make it sound better, make it deeper, slower etc. Must be something out there, there are plenty of phone scams.


----Mike Savad

Use your own voice, but speak with an accent, we can all do accents in some form…

I have been planning to create some marketing videos and gathering up all the tools I want to use. In researching the AI voice voice generators, I want to pay for a subscription that does allow for commercial use. I am wondering if YouTube and other host sites have a detection system that recognizes the use of non-licensed vs. licensed AI voice content.

Brian - Eleven Labs is at the top of my list of choices for text to speech generator and I planned on paying for the starter subscription which includes Commercial Licensing. Did you use the free version?

I wonder if google has their own version and this is yet another monopoly situation, or would they have it so you can use it.


----Mike Savad

Mike - Google does have Text-to-speech AI generator. Its free to try and use if you add a credit card, and the monthly limits before having to pay is pretty high. I don't know if a commercial license is included.

I do wonder if that google version will somehow be OK to use in this case... I'm betting it is.


----Mike Savad

I've just heard about this suggestion for families concerned about A I generating their voices.

Have a code word

Mike - You have a very good point there. It would be like slapping the hand that feeds you if it wasn't ok. I'm going to consider using it if I can figure out how to use it. Not everything with Google is user friendly.

Okay...so here's my follow up on your points/questions:


1) Why Would Using an AI Voice be "Against the Rules"? - it is NOT against the rules. After all, you can clone your own voice. Very very well. And "synthetic voices" have been being used for years. Without any problem. And half of all tick-tok videos use some kind of synthetic voice and they are all ported over into YouTube shorts. YouTube is in LOVE with shorts and is pushing them heavily so they are not going to make a policy that would have them nuke half of all the Youtube shorts in existence.

The devil is in the details. The main concern is what they call "synthetic content'. Which is code for, "Deep Fakes Need to be Disclosed".

So if you are going to use an AI voice to tell a story about a puppy who got lost and then was found by his little girl owner....there's really no need to disclose. But if you make a video where President Biden is saying he only loves dogs and hates cats....that needs to be disclosed. It's not even "banned". That would be political satire. There's effectively no difference "in the content" in that than making a cartoon of Biden saying he only loves dogs and hates cats. The problem is that with deep fake voice/video people will BELIEVE that Biden really DID say he hates cats. The cartoon format "auto discloses"....while deep fake humor content does not.

So yeah...you CAN use AI voices. That's not the problem.

The problem, as I see it, is that ANYONE can use the voices. And so if someone uses the voice to go on a "spam video spree" then the YouTube algos can start looking for that particular voice signature assuming that any channel that fires up and uses that voice "must belong" to the spammer who they banned months ago.

It's like the Alex Jones thing. They banned Alex Jones and then if anyone interviewed him....the algos would nuke their videos. Well...with millions and millions of hours of video uploaded each day...how could they identify that Alex Jones was appearing burried in the middle of a video?

Because your voice has a "fingerprint" and they just coded the algo to look for the fingerprint of Alex Jones voice. Same here. They coded the algo to look for the "fingerprint" of the voice I was using from Eleven Labs. I'm guessing that ANYONE who uses that voice will find their channel "accidentally terminated". Which...of course means that if you use ANY voice that is not solely under your control to make voiceover content...you always run the risk that someone else will use that voice in bad faith.


2) Couldn't someone Steal Your Voice - yes. They could "steal your voice". And that may be a problem in the near future. But I don't think it's a real thing now. Scammers and spammers take the path of least resistance. Dozens of voices are being uploaded to the Elven Labs "library of voices" per day...so why would you bother trying to take the time to download the audio from some random YouTuber's channel and then run it through an ai modeling process when you can just... PICK ONE that is ready to go from the Eleven Labs. They're like "burner phones"....but "burner voices" instead. And there's new "burner voices" being uploaded on demand every day.


3) Am I going to redo the videos with AI Voices? - well...yes. And no. I'm just going to bite the bullet and us my own AI generated voice. Eleven Labs has two ways to do that. They have a "quick voice" trainer where you just talk into the microphone for 30 seconds and it makes a mini-clone of your own voice that it can use. It's actually pretty impressive. But not as impressive as...

The FULL VOICE CLONE. In this scenario you get out a pro-grade microphone and record 3+ hours of audio. You then upload that file to Eleven Labs and then they take about 3 to 6 weeks and will do up a SUPER MODEL of your voice. It's really over-the-top amazing. I did that last year. But I have only used it in a test. Because.....

I don't like the sound of my own voice.

It really is flawless and sounds flawlessly like my own voice. So flawlessly that I dislike it. But I'm just going to suck it up and use it. Because that's the only "safe" path on YouTube I think. Use your own voice.

I'm going to watch a bunch of YouTube videos on how to improve your voice and speech and stuff and will practice and then do up a new 3-hour file and send it in to them. I might do it every 6 weeks so I can improve my voice inflection and get that out of the model. Also...as I get older....my voice in the model will match my age over time. It would be weird to be 70 years old and be using your 40-year-old version of your own voice.


4) Licensed vs Non-Licensed Use - my guess is that at this time there is no way for services like Eleven Labs to know if the people who should be paying are choosing NOT to pay. For now. My guess is that they'll will come up with a system like Epidemic Sound. You can use their music by ripping it from some other YouTuber and they won't give you a copyright strike. But they WILL "copyright claim" the ad revenue on your video.

You have have to pay the yearly license fee to Epidemic Sound and then add your channel ID to your back panel "white list" and then the copyright claims don't happen. However Epidemic Sound does this...I expect Eleven Labs and other such services might do the same.

But they may not need to. The problem is the quota. I burned through 100,000 characters very very fast and had to upgrade to the "next level" paid plan. Some paragraphs it just....voices it wrong. Especially fictional dialogue between characters. You might have to re-roll it 3 times to get it to sound right. That eats up quota fast.

So just usage likely will grind out the people who want to use it for free. It runs out too quick.

"2) Couldn't someone Steal Your Voice - yes. They could "steal your voice". And that may be a problem in the near future. But I don't think it's a real thing now."

There are reports of this happening already. Here are two news stories on it. (If you Google it, you'll find more.)

https://www.foxbusiness.com/technology/ai-voice-cloning-scams-on-rise-expert-warns
https://www.cbsnews.com/news/scammers-ai-mimic-voices-loved-ones-in-distress/

YouTube just announced its new approach to AI content in November, stating that changes would be coming in the months ahead. Things are changing, and that makes it hard to predict exactly what will be flagged and how the algorithms will work. It looks like one new requirements is that AI content needs to be labeled, but I'm not sure if this has rolled out yet. Did you label the voice as AI, or were you given that option?

Fascinating. Since I decided to try to join the 1990s ( ok I know it's really the 2020s)

....Since last summer or fall when I signed up for FAA and created a GoFundMe fundraiser and started looking at Kickstarter as possible fundraiser for creative projects I have considered YouTube. Especially since Kickstarter either requires or encourages videos and a friend explaining how Kickstarter links to YouTube videos and to have a Kickstarter video you must put it on YouTube so that when they click on your video it connects to YouTube. So I have some video ready and some ideas for several projects.

I haven't yet created a YouTube channel and already planned to keep my face off but now it seems I have to think about voice theft? To tell the truth I was already afraid to join YouTube....some reasons probably legitimate ... some fear of the unknown..(trying to figure out how to be rich and infamous artist and unknown at the same time?)

I'm thinking maybe me in a full mask using Sign Language with captions? But then anyone who is blind would be left out unless they have a caption reader on their computer? (This paragraph is not a joke in case it's hard to tell. Accessibility in general is one of my artistic themes and a field I used to work in for poverty wages.)

While I don't need to hear my own voice either and will have to think about this...though my first thought is you can't speak publicly unless you use your own voice (or a fake accent) and my voice has been on radio and TV over the years so I'll probably continue to use my own voice moving forward.

Brien......I like seeing my photos or my face on TV less than I like hearing my own voice.......note my "Don't Hate Me Because I'm Beautiful" artist photo.

You might consider just breaking down and using your own voice. With every sentence and hour you record your own work you will improve. Though imagine someone stealing your voice and calling you up....."hello....is this me?"

I'm not ready for all of this and yet here it comes!

https://arstechnica.com/information-technology/2023/09/ai-generated-books-force-amazon-to-cap-ebook-publications-to-3-per-day/

Perhaps they are facing the same issue that Amazon and many POD sites are. AI just makes it too easy to make content and they are worried viewers will be put off by the sheer volume of AI clips they need to wade through. Good to see they are addressing this, athough in your particular case the regulations seem too heavy handed. I understand why Youtube feel they need to somehow manage AI content though.


I would just use my real voice and not worry too much about how it sounds. The human element might actually work in your favour.

Google and FB are in overkill over stopping posts and ads. FB offers the better chance of overcoming the hurdles and letting people know what is going on.

It is important not to assume why. That is hard. Worse Google won't explain it at all. They say that in the ads department.

This is on FB ads. It is three years old. The only thing that has changed every ad is now challenged and later on someone will see it and allow the ad.

I get you are only posting but your posts are meant as ads. Google is worse on this.

FB quarterly ad revenue was up 26%, while Google's was up 11%. I see why people can not work as they wish with Google.

This may seen paradoxical but Google's and FB's challenges are meaningless. FB I feel comfortable getting the first challenge overturned. Google not at all.

Off-topic, but you may find this interesting David

https://www.theguardian.com/business/2024/feb/15/x-paid-post-promotion-advertising

I won't touch AI for my videos, I won't even use chatgpt for the descriptions. YouTube will probably be cracking down hard on AI generated videos because automatically generated low effort AI videos are flooding the platform. Even if you don't fall under that description if an element of your video has AI in it they might do what they did to your channels (which sucks, sorry to hear that) or deny monetization.

Text to Speech audio has been used for years for making videos. It's not the problem unless said voices have restricted usage agreements and there's breaches to said agreements.

The massive volume of unregulated unverified AI generative stuff being dumped online is what is in question.

YouTube is a business. It has the resources to separate the chaff from the grain.

Other online businesses also need to find ways to protect the intellectual properties of artists as guaranteed by the US Constitution not only because of ethics but also to mitigate potential class action suits.

Think of natural gems with all their flaws and the way these flaws are like fingerprints: unique identifiers.
Lab created gems have chemical tracers to keep the trade of gems lucrative. Tracers also should be mandated within all generative AI creations.

Human artists need heed lessons from these artistic jewelers and continue to demand mandated identifiers to keep human creations lucrative.

I did choose just to create a different channel with my own voice. Now begins the multi-week work up process.

I did find out that if you load the audio file into premier…you can open the “Essential Sound” window, label the audio as “Dialogue” and then go in and apply a podcast sound to it.

Then you export it and if you Google “Adobe Podcast Ai”…they have an online app you can upload the file to which essentially acts as a full-blown sound engineer.

Using this two-step process the audio is coming out flawless. It even removes the sounds of the fan from my computer and the cars driving by outside.

I do not touch AI because I do not own it.

We have enough infringement.

It is not an infringement in AI cases. We have no recourse.

They really do make everything into a pile of work don't they?
For what it's worth, my advice is to take some voice lessons, get used to your own voice and ditch AI altogether .
I block pretty much every video channel that uses them, and at least at this point, I can always tell.
They lack humanity and expression, and as a trained singer, they annoy me to no end Again, though, that is just my personal feeling on them. You certainly aren't required to take it as law.

Wow! What an ordeal! I'm always afraid about using music on my videos because of copyrights. Even though I've downloaded them from the free music library, I've had a few that popped up copyright. One was taken down and the others just have a copyright not but says copyright owner allows it on YT so it's not a flag. So anymore, I pretty much just keep reusing the music I have that I know is safe even though I'd rather have new music. I tried to download some yesterday for a RE video for a client, but it told me I had to upgrade to the YT premium account you have to pay for and didn't really want to do that.

deleted

Isnt there a A! voice generator that allows you to manipulate your own voice to improve the tone and quality that you can save as a default voice for all your future videos?