Here I will condense some ideas I've written about in other parts of the series; to make them more clear, and also to serve as a kind of introduction to people who have not read them. a true TL;DR thing.
When creating music with AI, there are of course several ways to go on about this.
Here are two of them.
1. Something I call the "one-click method" (even though it's rather typing than clicking).
People are used to apps like AI art generators, where they can type in one prompt, the AI generates a complete, finished picture.
Now they are looking for something similar in the world of music: a way to type in one prompt, or even several prompts, and the AI outputs a complete, finished track or song. Everything is automatized, everything is "ready for release". For example, typing in 'write an epic rock song with a powerful chorus', and the AI does just that, perfectly, and everyone is happy. Making a full track with the power of "one mouse click", so to say, metaphorically (even though it's rather typing in... well, you know that by now). Just like you can start a program, download a file, etc. just by one mouse click.
And if not a full .wav file output - at least a "full midi export" that you can just load into your daw, or something similar.
Many people are following this method, are developing apps that should work this way, and so on.
But, so far, I wasn't very happy with the results I heard, and there are lots of problems and obstacles on this route.
Maybe this will work in the future - but i do not think it works very good right now.
2. what i call the "human intermediary" method.
the idea here is that instead of aiming for 100% automation in AI music creation - the human performs some tasks in the production of music, too!
in it's "strongest" approach, the AI becomes the band and the executive music producer, with all the creativity, ideas, vocals, melodies, harmonies, rhythms, structures, and so on, the whole of the track or song - and the "human user" becomes the subordinate mixing engineer, who has to perform a lot of manual tasks.
the human becomes a kind of connection or bridge, between the AI, and the DAW or other equipment the music is produced on.
in a "weaker" approach, the user and the AI share (some of) the creativity of producing a track, and collaborate on these things.
there are a myriad of ways to go on about this; i outlined some of them already.
but in one way or another, they evolve around the same idea: the ai generates melodies, notes, patterns, or the outline, structure and theme of a track (and everything else you want) and the human user has to input this data into his DAW to generate the track.
if you have not done this before, this might be a bit hard to envision, so i've written a simple step by step example for creating an oldschool acid techno track this way here: https://laibyrinth.blogspot.com/2023/08/how-to-write-music-using-chatgpt-part-2.html
And a more complex (but still step-by-step) one for creating a 90s style hardcore techno track: https://laibyrinth.blogspot.com/2023/09/how-to-write-music-with-chatgpt-part-4.html
the possibilities are endless. the ai can generate any aspect of your track, or even brainstorm a track when you have run out of ideas, etc.
while most people look at method 1 right now, i think the "human intermediary" approach is more promising right now - and maybe even for the future.
some afterthoughts on this. (and it's not tl;dr anymore!)
i think an issue here is that the world of ai music right now is populated ai tech and diy music enthusiasts (don't worry, i'm one of them, too) that have a certain conception about music productions. they want a "jinn" that outputs final tracks.
but that's not how music production (without ai!) usually works in the "real world". far from it.
for example, there are big record labels that are looking out for artists, in hope that these artists become big stars and generate a ton of revenue.
they are looking for people that create extraordinary music, have special talents, maybe also a special voice or performing ability. that "extra" in creativity that normal people lack.
they will sign these people and spread their music all over the world.
*but*. they are not looking for artists that bring them finished songs or tracks! artists that come to their office and say, "hey, i (or my band) produced this album, it's very good music, it's finished, just upload it to the streaming services".
it's enough if the artist has *written* the songs, in the sense of having the notes, chords, vocals, and so on.
*finishing*, actually producing the track is then the task of the producers and mixing engineers in the studio.
(there are some examples of famous bands who actually *produced their own music* - but this is the exception to the rule, especially at the beginning of a band's career!).
and this is where AIs like ChatGPT come in.
they might not give you output in form of a final, finished track yet.
but they can take the role of artists, who write tracks and songs, i.e. writing the notes, verses, chorus, chord progression, and so on.
and the AIs can give you that extra creativity, that sonic genius, that is so rare in this world. and maybe even some creativity that is beyond existing human levels yet!
*but* you, as the "subordinate human mixing engineer", still have to do the manual task of actually producing the track of the AI, to get a truly tangible result. at least for now.
No comments:
Post a Comment