Home Creators Posts Import Register Favorites Logout

Content

Hello guys, I would like to share my experience with you that I recently made with AI generated music. So far I actually rather resisted using AI in general and haven't used any ChatGPT or other LLMs, although I had already used some free image generators before as inspiration for logo designs etc. but until recently I have rather avoided trying out any AI music generators. With all the AI music popping up everywhere I finally I couldn't quite resist and decided to fool around with it on my own a bit. Meanwhile I am both fascinated and shocked with the results and would like to show some tracks that evolved out of my experiments and also talk about the issues I have with AI generated music.

All tracks I generate will be available to the enthusiast level free to use for whatever purpose you want and I will also offer collections of tracks that may somehow fit together as "albums" on Bandcamp. But the main purpose is to use this music for our public Youtube videos. We already used AI generated tracks in our most recent videos on Youtube as some of you may have noticed. All tracks in this video were created by me or my son. With my old Oluvsgadgets channel I always had to fight with copyright and many videos got demonetized after some time because suddenly they got some copyright claim which at first didn't seem to cause any issues. Therefore I had subscriptions for Epidemic and Artlist, which also cost a lot and usually are a mess to find really decent sounding and musically interesting tracks. I think the money is better invested into a Suno Pro subscription which lets me generate the music I want and which will hopefully avoid and copyright issues in future. For Patreon nothing will change, I will still try to select commercial high quality tracks which are suitable for demoing the sound quality of speakers headphones etc. because the biggest problem with AI generated music is that it simply doesn't sound perfect and contains lots of digital artefacts which are not that easy to hide. Raw tracks created with AI require a lot of post processing and several mastering stages while may still appear a bit "fake" and digital sounding in terms of recording quality, not in terms of "music". Actually the musical quality Suno can create is quite astonishing and if used professionally it can be a huge inspiration for any musician even if only for finding some ideas.

I had stopped creating my own music about 25 years ago simply because it was very time consuming to achieve the results I really had in mind and my setup back then was always quite limited, this was long before having a full "studio" within a Macbook with just some USB master keyboard attached. When I started to study I quickly noticed that I couldn't invest that much time into creating music besides I always found myself musically not talented enough to really being able getting some benefit out if it in the longer term.

With AI it's a completely different story as it can accelerate the workflow a lot, not by creating completely new tracks from scratch but rather by using it for ideas and inspiration. A lot of today's music is completely generated on computers with no real physical instrument being played at all, so downloading single MIDI stems of some bass line or arpeggio and using them for your own instruments or plug-ins should result in a much more professional sounding music in the end than an entire track fully generated by AI. I know that many music producers already use Suno since the beginning, and honestly I am shocked with the outcome hearing sounds or hooks and basslines I would have had a hard time to create on my own.

So far I mostly used my own tracks that I made in my youth 30 years back and wanted to hear how the AI would interpret or "cover" them into a more modern and more commercial style that would also make me want to listen to them again. As my own music was also created with Synthesizers and my intention was always to make it sound as electronic and as synthetic as possible (I grew up with Jean Michel Jarre) I have no problem with the outcome from Suno sounding electronic as well, I actually like the way how Suno mixes and uses synthesizer sounds, renders electronic basslines etc. it would be definitely different if my intention was to create "natural" sounding music with real instruments, here it would be way harder to achieve a really convincing outcome. Creating tracks completely out of nothing with Suno can be quite random at times, therefore I rather concentrated on using my own music within Suno and make it creating something different and often even better than my original.

I am not sure if anyone is interested in this topic at all, but music was always a huge part of my life and now suddenly being able to hear my own music in a new form and way that I was not able to achieve back then makes me definitely baffled but also sad, as I never really managed what Suno does within some seconds. Of course not every attempt is usable, the outcome may be quite random at times and it may need several approaches, maybe also remixes and additional edits to make it sound like I have in mind. Sometimes Suno simply ignores the prompts and adds some vocals although I explicitly tell it no to use any vocals. Therefore there is lots of trial and error involved if you have a precise vision of how something should sound.

For the start I would like to share a cover that Suno made out of my own track "Better Times" here an excerpt of my original version with the Vocoder-solo part:

This version from Suno stays quite close to the general structure of my original song, but interprets it slightly differently and more modern and in some cases much better than I managed, although my solo is completely missing here. You will maybe also notice that the raw version has some audible digital sizzle which is not quite possible to correct for:

And here my mastered version which still may need some work to make it sound "perfect", I also had to apply some dithering and some analog tape simulations in order to make it sound a bit more natural:

Another example of the same song used as base which shows that AI is not that impeccable either, I mean this version has its moments, but is definitely something quite a bit different and the overall "sound" is also quite off, but honestly I have already heard worse music being released. My favorite part starts at around 1:01, it even has a bit of a Herbie Hancock vibe...

If there is interest I may do some live screen capture sessions of track creations directly within Suno while commenting on the outcome. Just let me know what you think about it.

I hope this was at least interesting and I would encourage anyone interested into music giving Suno at least a try.

Files

Comments

Mika Eel

I would love to see you make it live. I think it is a very interesting topic. I also used udio.com for creating, which was great experience

Mika Eel

Wow this track is really good. Dreamy, goes forward stomping hard! and has some nice distorted effects. The vocoder sounds awesome! Can you make it downloadable for us? As you said, its quite shocking (but not surprisisng) how good it gets that fast, and its shockingly only the beginning... Its sad, yes, that the craft of music making gets lost but it is and was always like that with technology. You still can make your music the old ways but as You said yourself it is time intensive. If i make a track using ai it also is timeintensive but in the end i am happy i gave instructions and worked o tweaking it for hours and sometimes days. I recently repaired my plasma tv, exchanging/ soldering transistors and capacitors and was fully guided by googles gemini ai, which is crazy accurate. You can use all of the premium LLMs without cost at arena.ai and you should always use two side by side to watch out for errors that one does and the other doesnt. I use gemini and claude models side by side on the arena site. You have to choose the "side by side" option and the models at the top (and dont put in personal informations, but i hope this goes without saying, as it is known that ai's are eating up your information ;)

J

I avoid ai as much as possible, i only support real artists

DominikReviews

I remember using Suno in the early days and it was okay but had very artifical vocals, lately tried it again and its much better now but still not somehow very good.

Domingo Martínez

I'm in the same boat. It's incredible the wonders Suno creates from my unfinished Cubase productions. I'm absolutely hooked. I think if they can improve the sound quality, it'll be unbeatable...

J

Probably the reason why i dont like the newest music then (2020+). Music is not the same as it was in my opinion and probably will never be the same due to ai and kpop bullshit...

oluvsgadgets

Have you tried outputting single stems as MIDI?

Domingo Martínez

Yes Oluv, of course, I transform the drums via MIDI with better, cleaner samples, but the bass, synth, etc. sounds are much more laborious since the expression and effects that Suno generates aren't reflected well in the MIDI tracks it creates, and doing it from scratch in my DAW is a pain.

oluvsgadgets

I also noticed that it does solos and bass filtering really nicely, it’s hard to do it that well. But mostly drums and especially hi-hats somehow sound weird from Suno. I would like to hear some of your tracks

oluvsgadgets

sounds great, although the intro is a bit long until it goes to the vocal part. And it's funny as Suno seems to have this typcial style, it also created tracks from mine that sound similar in terms of sound and instruments.