Most (old school) copyright detection algorithms search for exact matches to the copyrighted work, so by making it recognizable to humans but not an exact 1:1 match to the original work you can get it to be a false negative for copyright infringement.
However, I mentioned that it’s old school copyright detection because newer algorithms know the common tricks and will search for mirrored visuals and edited audio, and AI is making it increasingly difficult to use the slight edits. It’s why sometimes the video will be cropped *and* mirrored, and the audio will fluctuate between being too high and too low, to try and get around the more advanced algorithms.
Latest Answers