@Edges is using AI to generate music.
The basis of most AIs is knowledge. An AI has to know what to do in order to meet its given criteria when generating something. Take ChatGPT for example. A few years ago, it was not very good at its job and was more or less an internet toy for people to play with. Nowadays, ChatGPT’s writing can easily be mistaken for a human’s. This is a result of better training and just more data to learn from in general.
Let me reiterate. If an AI is badly trained / has not been trained very much yet, it won’t have the required knowledge to perform a requested task to human/near-human standards. It’s most obvious in AIs made to generate audio, video, or imagery. Every kind of AI has its own rendition of the 7-fingered hand.
Music AIs are relatively new, yet their giveaways are no more obvious now than that of when thispersondoesnotexist.com (for example) was introduced. People thought those faces looked real. Now, not so much.
Usually, an audio-based generative AI will create a white noise signal, then use subtractive spectral synthesis to pick out whatever frequencies it needs from the white noise to make a certain sound. Sometimes it fails to the latter step effectively. If it’s not 100% sure of what frequency to select at a certain point, some of that white noise will seep through.
It’s comparable to a person mumbling or stifling their speech when they’re unsure as to how to pronounce a word. At current, it’s audible in EVERY music AI. You can hear it in leads, chords, and pretty much everything that isn’t primarily a transient.
It’s audible in @Edges ‘s tracks. Where I said it would be. It might be hard to hear at first, but if you listen to a few audio samples of a voiceover AI screwing up that frequency picking thing I mentioned earlier, it gets more and more obvious when you listen to AI music as well. This track contains some examples. Neon Dreams has noisy vocals as well.
Create an account or Login to write a comment.
Its stupid to me, to think that using a bit of AI is dumb, I think sometimes with the creation of an idea in the right manner and original ideas on your end, AI is used correctly. However, clear as day using it as if you MADE the music is a diff story and I aside with these individuals. That is extremely unprofessional for this site and for those following, and creates worse trends than good.
Now I know what made them mofos so damn mad 🤣🤣
I go to practice and this what i miss 😂
you know what's still stupid about AI is that people still use like it's a plaything, especially with image generation
just goes to show how seriously we can take it if it's just a toy, but i will say it can be great for shitposting
I'm disgusted by this kind of people who make music just for the sake of making it and more with AI, the worst way to do something, they should be ashamed of themselves
exactly
that user was not an artist
Well, my conclusion is that the person who did that does not deserve to be called an "artist".
thank you lol it was really obvious from the start, glad something is being done
this track was a blessing
lol
why we targeting a person niggas
oh this kid lol
Yeah, this makes sense now. Especially when they commented and it never made much sense, no offense. I wonder if they use ai to write the comments to.
jesus 20 comments in 15 minutes
Republished
fixed BPM
Republished
Music AIs also have very obvious inconsistencies with stereo audio, compression, and quantization (staying on beat). You can, once again, hear it in this user's tracks. You can hear the source/position of various sounds shifting (stereo inaccuracy), very bumpy compression, and other things like mixing inconsistencies (a snare will have more low end on one hit than another, or be quieter than the last hit, and those differences will not occur in a pattern like expected if it were created by a human being)
I ain't reading all that
