Lately I’ve seen a lot of AI remakes of popular songs in completely different styles — for example on YouTube channels like Fake music, Imagine, or almost real. I’m really curious how these are made. Are they using something like Suno Studio, or maybe a combination of stem separation and voice conversion tools? As far as I know, platforms like Suno or Udio don’t allow you to upload copyrighted songs or lyrics.

Does anyone know what software or workflow they might be using?

A few examples:

 

Basic principle: Why AI and not traditional production?

AI makes it possible to transform songs quickly and cost-effectively, without the need for an entire studio or musicians. The process often begins with the original song, which is disassembled and reassembled. Platforms like Suno or Udio actually prohibit the direct upload of copyrighted content to avoid legal issues. Instead, creators use workarounds:

They generate new elements based on descriptions (prompts) or edit local files with open-source tools. The end result is often "inspired by" the original, but not identical – which places it in a gray area. Many channels monetize this through views, but there are criticisms that it displaces real artists or leads to fake content.

 

i'm not sure about these specific ones, but I watched about 5 minutes of a tutorial video that was supposed to be about finding good vocal samples. The guy started talking about getting vocals from Youtube videos and Soundcloud with the specific search string of "isolated vocals". I looked on Soundcloud for the first time in awhile; and there were thousands of tracks of this junk on there, "good" or not. I turned off the tutorial and that was the end of that. They are obviously AI generated somehow 5+ of every one, as you have noticed.

no thanks.