I haven’t yet watched the video, but from the posted images, it looks like they converted the image itself to a frequency spectrum, so no image format encoding was involved before it reached the point of being audio.