AI Voice from ElevenLabs

Hello Everyone,

I am planning to launch my first course on Udemy. As I have strong accent that works badly for my overall course quality, I decided to use Text-To-Speech model for my voice. I have gone through many tools including teaching my own model on open source voices, but it was quite robotic and I saw Udemy is not accepting the courses if it sounds like it was generated.

The best I could find was latest model from ElevenLabs, I tested it (with "Liam" voice if someone is interested), and the audio quality sounds nearly perfect. Even though I am not able to spot the difference between this and normal human voice, the reviewer of my test video noticed it, and told me that there might be issues when I launch the course and I should consider getting better TTS model, but as far as I was doing the research there is no better one.

Did someone of you pass the course review with ElevenLabs voices?

If not, what other models can you recommend that would pass the checks?

Comments

  • ElevenLabs should be fine. Just to be sure - create one lesson and send it to Udemy as a "test video" before you launch a mass video production.

  • The large problem with robo-voices (robo-voices is an old 1980s technology) is in the realm of copyright and that the current legal landscape clearly makes Udemy liable for copyright issues related to both the ownership of robo-voices and the ownership of whatever materials the robo-voice reads and records.

    The current interpretation of the Udemy rules for use of robo-voices *seems* to be that you may may make your own voice into your own robo-voice and use it to create your own teaching materials on Udemy for the purpose of removing injuries, disabilities, sickness, and other voice related issues.

    When you use someone else's robo-voice you lose the copyright and create a huge uncertainty with respect to the complete copyright of the material. For example it is common that new young academics simply reads from some old "stolen" book or teaching material and gets removed for copyright violations. Using a robo-voice in the same fashion would in principle turn Udemy into a "pirate bay 2.0", with lifetime purchased access. It is impossible to conceive that Udemy would accept that in any way.

    Concerning Elevenlabs, I tried it about 6-12 months ago. The socket-dolls are clearly robotic and the "make your own robo-voice function" totally sucks minus five stars. In the Elevenlabs case their problems can clearly be derived as being due to a too small training dataset and the use of inferior model technology.

    You shouldn't pass a course review with a Elevenlabs socket-doll, and even if you managed to confound the review and passed, that wouldn't give you a green card for future socket-doll endeavors or the right to keep the materials that passed in the market place.

    Regards

  • @HenrikJohan618 Did you try to use ElevenLabs and sent it for a final review? All I got from my reviewer for the test video was:

    "The AI voice doesn't sound very natural, if possible I would suggest using a different AI voice that sounds more natural. There are a lot of AI voice-over options, so I would suggest checking reviews and samples online before picking one. Also, as a tip, I suggest using an AI voice that resembles or mimics your voice, as much as possible, to prevent identity issues when submitting the finished course and to provide uniqueness to your course."

    Which is not giving me straight answer that "yes your video will be accepted", or "no it will not work". When I am supposed to spend nearly 200$ on the voice I would like to be sure that it will pass.

  • I am just thinking what would be a better idea, to go this AI voice way, or to buy a microphone and to learn to speak :D

  • @Coffinonv

    I don't use AI voices at all in my production but was at that time looking for an AI-voice to fill in for various sound kills due to my booming voice and the fact that a lot of microphone and recording hardware/software are trained on Taiwanese voices and therefore auto-creates various sound artifacts in the recordings or removes sound with or without replacement.

    I will offer you this advice, what the Udemy message means is that you must create an AI-voice that sounds like your own voice for (implicitly) legal and contractual reasons. The other words just means that the robo-voice making software will map your sound to another persons voice and "put your sound" on top of their speech so that they speak with your voice, not that you should use their voice instead of your voice.

    All other things is your private business alone.

    Bye

  • Even if you are using robo voice or your own sounding Ai voice, how do you get the talking head? And if it is not there is any way you would not get the rating to get your course going. I have a very hard accent, but I still do not use the robo voice or AI voice as I could not find a practical solution to provide a good experience to the student using these tools in my courses. Any comments? Am I missing any new technological breakthroughs?

  • Hi @VijeshJain - it's all about your vision and definition of a good course or video. Why do you need a "talking head" in the first place?

    Watch some nice faceless YouTube explainer videos - most of them are done with AI voiceovers. I understand that some instructors are trying to build a personal brand here, and they center their teachings around their own "talking heads". And, that's much easier for native speakers - for the rest of us, especially those with strong accents - well, we have to find our own "business model", not to copy them.

    After all, content is the king, and that's what students are looking for. Just check "365 Careers" among Udemy instructors - these guys made a huge business on the platform. They rely on "2D animations", while you can find some other instructors relying on "Whiteboard animations".

  • Thank you all for the reply, I will borrow the professional microphone from my friend and let's see, maybe I will be able to fix my voice with audacity, or just learn to speak better.

  • Yes; try with your voice, enhance it with audacity and clean it from wrong words or bluff and send it to Udemy test video to get review.