If you would have asked me which one is harder for A.I. to solve in creative space between images and music I would have been wrong. While the image generators have been out for longer already they don't really meet the quality that the music generators can output.
Might be just me, but for the music I can get satisfactory results from the A.I. but for images there is always something wrong and they don't really meer my expectations.
It's also interesting that the music generation services are much cheaper than the image ones. Even though they both practically work the same (as music generators also generate images of the sounds' waveforms) the music "images" take less computing power to generate as they aren't necessary needing as high resolution.
On the other hand producing those outputs by comission the costs are quite the opposite. You could comission a picture from an artist quite cheaply, but producing an album (or just a single song) could cost vastly more and wouldn't really be an option for having your songs produced just for fun.