| Great. I feel like the music AIs are flying under the radar at the moment. I assume they will pop off shortly, once people realise how good they have become.
Hopefully they add an upscaling option for songs v<1.5. Related thoughts - - I struggled with Udio to get great results. This might be my poor prompting though. - Suno was better for me - so far, and my prompting is far more successful. This Udio update will hopefully bring it into line with Suno. - I have had quite a few 'HolySh__t!' moments with Suno, and can now often induce frisson with it. - More and more, I'm referring to my own creations rather than my Spotify weekly suggestions and Spotify playlists. Probably 1:1:1 ratio now. - Listening is starting to merge with creating, i.e. I'm listening, extending the song, choosing the best extension, and so on. - I have a 3 part, 15 minute song that I am actually in awe of. Possibly just my simple brain that loves beeps and boops in my ears, rather than it being a work of art. - I suspect I'm a layman musically and that some people would find some of the transitions and patterns formulaic / refer to pre-existing art, but for me it is firmly in the 'wow' stage now, and it can fairly reliably garden-path my brain. - Generative music like this is making me relate music to comedy - Music can almost be like a joke where a slight adjustment in the song tricks the brain slightly and can trigger really interesting emotion, ranging from "ah of course!", "genius!" to "oh that's nice, I wasn't expecting that" and "like a key in a lock". - The above, for me and for now at least, is only for instrumental songs. Lyrics and vocals still seem banal, robotic, boring, cookie cutter. I haven't managed to get any good lyrics on songs other than a bit of techno with intentionally repetitive/robotic background vocals. - Suno's extend option for tracks seemed broken a while back but is now fixed. - I've been thinking about how this affects the live music industry, and I suspect that the DJ scene will be partially eaten from beneath __soon__, once on-the-fly generation is launched - i.e. 'why should we hire a dj for our bar / party when we can just ask SunoDJ to do it for us'. - Other than on the fly generation, what I _really_ want is for the tracks to be split into stems. - Suno's paid tiers = you own the copyright. Suno's free tier = They own it. Not sure about Udio. - Suno = a constant battle to keep it innovative. It does often go generic in terms of timbres, structure etc., but you can break out of that. I think. Again, it might be generic crap I'm creating, and it's my illiteracy that makes it seem unique. - Dub/dancehall/reggae and MelodicTechno/AmbientTechno are the easier genres to wrangle IMO/IME. |