Recently, I started uploading videos to Youtube using AI voiceovers. At first I was uploading mute tutorials, but someone complained that I had to explain what I was doing. The reason I wasn't doing that was that I didn't want to put my own voice in the video. But then I realized I could just generate the audio with a text-to-speech software, so I started doing that. There is only one problem: it's "AI."
I'm using an open source TTS with an open-weight AI model. but Youtube only has one checkbox for AI-generated content, and it discriminates as follows:
- If you cloned your own voice, you don't need to check it.
- If you cloned someone else's voice, you need to check it.
This isn't my own voice, so naturally I need to check it, which is kind of awkward.
Keep in mind that my videos are just:
- A recording of the desktop with someone narrating what I'm doing.
I need to check the same box that I would need to check if I were creating content such as:
- A "deep fake" with AI-generated video making a celebrity say something they didn't say.
- "AI slop" featuring AI-generated images with AI-generated audio narrating AI-generated text made almost entirely with ChatGPT.
I'm not convinced that a single checkbox is sufficient to group these three use cases. Personally, I feel like the problem with "AI" in this case is one of authenticity, and we can categorize the inauthenticity of videos featuring "synthetic" content in three ways:
1: fraud: videos that present something that didn't happen as if it did. These are dangerous because they can cause harm, and it's probably the type of AI-generated video that everybody would agree should be banned.
2: cheap: this is the type of AI-generated content that offends viewers because it feels like someone is just creating "content" so they can get ad money from the viewers' attention. A single video like this doesn't have the same potential for harm as a fraud-type video, but because this type of AI-generated video is so easily to mass-produce, there is a chance it causes harm for the ecosystem where videos are published, e.g. by filling the entire platform with AI slop.
3: utility: in this type the AI-generated content is being employed in the video and doesn't mislead the audience in any way about the content of the video, so most viewers wouldn't mind it—although there may still be some who simply dislike AI in general, but I don't think it's possible to reason with someone who can't make concessions.
Although I feel it's very easy to draw the line between what is fraud and what isn't, I have to concede that it's unfortunately not as easy to draw the line between "AI slop" and "not AI slop." I don't like "AI slop" as much as anyone else, but I think if someone is making the videos and people are deliberately watching them instead of human-made content, then it must be for some reason.
Some people may be generating the whole thing with ChatGPT, but there could be someone who is trying to tell a illustrated story they wrote themselves, and they want images but they don't want to pay an artist or make a deal with an artist. In this case, if they didn't use AI, they wouldn't make the story, so the options are either AI or nothing. This is just like if I didn't use an AI voiceover, I wouldn't make a narrated tutorial. I could try to find someone to do the voiceover for me, but I don't want to.
A better question to me is whether they should be allowed to monetize AI-generated images, specially when you have cases like ChatGPT stealing the style of a famous Japanese animation studio, Ghibli, and that incident becoming viral this year.
How does this question affect me? I use an open-weight model "trained exclusively on permissive/non-copyrighted audio data and IPA phoneme labels" [https://huggingface.co/hexgrad/Kokoro-82M#training-details] (accessed 2025-08-31). I have the license to put this audio in my videos. It's different from someone using ChatGPT trained on copyrighted images from the Internet to generate images for their videos.
I don't want to sound like I'm saying "it's different when I do it," but it IS different the way I'm doing it.