New Musical Journey – Part 06

Sometimes I feel like I’m being absurd and inconsistent in what I’m saying… And what causes this feeling? The fact that I’m able to talk about one thing and deviate for an entire article before coming back to it in the next one. So I wanted to tell you about the twists and turns of my life among the stars (as the late Corbier sang in “La galère capitaine” which I recommend)… Or rather my misadventures as a creator of original content with artificial intelligence but before we get there, I wanted to show you how we “work” with the raw material.

And then, for once that I open my door during the creative phase and give explanations that could be of some interest, you could still pretend to be interested, you bunch of slobs… Oh, yes, really? You’re starting to like it because you see a little more concrete but you find it a shame that until now I haven’t made anyone listen to anything I’ve generated? You’re right: we’re going to do a bit of listening compared to what we have in the alternative versions in this article.

Before that, we’re going to talk a bit more about technique and organization because it’s going to be important later on and the sooner we understand the mechanics, the sooner we can get familiar with them and not mess up… Because yes, there’s room for messing up and believe me it’s fast! So it’s not always very serious but sometimes we can still waste a lot of time… We’ve already discussed the “This is an Agglo” syndrome (generations of more than 25 minutes) and we’re going to see the other bullshit to avoid with SUNO and the output format issues… Because yes, we’re not going to hurt ourselves for nothing!

And if I still have a bit of time, I’ll tell you more about the rest of my journey in the music world. Reminder: I don’t know anything about current circuits because for 15 years or more I lived in a cave with my guitar, my PC and a not very clean internet connection. And then with a total aversion to social networks, it’s not easy to be interested in them and follow the content… However, I could have taken the turn at the time and become a YouTuber from the start since I already had a series of videos for “The Odyssey of Stupidity” in which we had chapters like “smoking kills, but not enough yet”.

In short, we’re straying from the subject but it’s to say that social networks and the rest are still a discovery for me in 2024!

Yeah so you’re going to love it when I tell you that AI is able to interpret what we give it as a sample to eat and that it manages to self-destruct a track that has a bit of breath on it. As I may have already mentioned in the previous article, when I recorded my first 3 “PrEvIoUsLy UnReLeAsEd” I did it with the means at hand and it was more with the dick and the knife than anything else and again, the knife was optional and I don’t know if you’ve ever tried to record with just what’s left but let’s say that if your name is not Rocco Siffredi you’ll have a hard time triggering the recording! More seriously, at the time my recording equipment was limited to its simplest expression and inevitably the demos have a very special sound. What I suggest is to take one like that cold and see what it looks like originally. We’ll take “Cathedral” to give it a try. Here’s the original demo:

It clearly states “crappy sound from the original demo” on it… It’s a Controlled Designation of Origin!

Well, then obviously, when it’s like that you have to start by treating the sample bit to balance it a little: first step, we run a compressor on it to eliminate most of the saturation and if that’s not enough we reduce by 2 decibels using our friend AUDACITY and its onboard effects which will allow another technical feat: balancing the bass and treble. Because yes, in “Cathedral” we have the bass which saturates to death and the guitar acts as an asthmatic mosquito which painfully agonizes in the treble. We therefore have a possibility in AUDACITY to modify the sound a little for our sample, which will give more or less this:

The same reworked…

So despite all the tests and tinkering that can be done with the effects, when the saturation is so “one” with the original demo and that we only have that to launch to feed the machine, well already it’s not brilliant but then it poses some problems because as I already said previously the artificial intelligence will analyze the song and for it it will be very clear: the saturation is part of the song and therefore it will be necessary to reproduce saturation, even if it completely screws up the generation after! I have some examples of tracks that ended up in “destroy” mode like that and when you listen to some of them you have the impression that the AI ​​ate the version to then vomit it up and eat it again like a certain Steve-O in a show called Jackass… I admit, I looked for the *.gif on Tenor.com and I didn’t find it (fortunately). So, this is what you get from audio vomit puree when you start from a demo with a sound that is a bit too dirty:

A complete generation with a disgusting sound at the entrance!

So first comment is that the input as I give it to him is slightly modified and even more disgusting than originally (yes yes, it’s possible). So why? Simply because my input format is compressed in MP3 and even if it’s 128 Kbit/s or even 192 Kbit/s there is still a slight loss of quality that SUNO will amplify a little… That plus the saturation, we are almost certain to have an even more disgusting track at the output… And yet, the version of “Cathedral” is awesome! It uses the sound of the sample well to build the rest of the song and with the lyrics it works but damn saturation kills your ears! Arf, it’s awful but it illustrates a well-known law in computer science that can be applied in everyday life: “when you give crap as input, you shouldn’t expect to get anything but crap as output… otherwise it’s alchemy.”

When AI starts messing up during generation!

More and more often I am confronted with this phenomenon of AI running wild in SUNO… It starts a track quietly with a correct and listenable sound without the slightest breath and the slightest saturation and suddenly it generates an oversaturated and messy track in which there are lots of things in superposition, a real audio chaos in which we no longer understand anything. Some will say that by choosing exotic sounds between “doom metal” and “sludge metal” or even both together with a “Dark” option we can easily obtain this kind of thing but I still have a hard time explaining it, especially when the previous part or even the previous parts are clean! We should not take the children of the good Lord for wild ducks, no way! In short, it consumes 5 credits for nothing and it is unusable…

When AI starts adding improbable things at the end of the generation…

There is also the height of incomprehension for me. You generated your track and at the end you find yourself with people talking, telling unintelligible nonsense or, like here, a psychopath having an anxiety attack when in reality… Well “what the fuck” what? And if it doesn’t end in an asthma attack in real time we’re lucky… Call the ambulance, I have a guy who dies while generating my track! No but seriously, you have to expect that this will happen from time to time and in this case be merciless: if you like the beginning and you’ve been thrown some out-of-context crap at the end of the track that ruins your recording, the only way to recover the track is to do [Extend] from the last place where the sound is good. It will cost you 10 credits but at least the beginning of the track will be usable. If by some misfortune it happens at the beginning of the track, throw it in the trash and pretend it never happened. Some say that these are “leftovers” from the training of the AI ​​model that come out from time to time. Do we have a ghost in the shell?

…or to scam us over the duration of the generation!

Another more than frustrating but extremely rare phenomenon: the Made in SUNO mockery or “what if we generated a 1 minute and 49 second track just to say that we almost got close to two minutes for the generation but we left like 1 minute and 4 seconds of blank space in the track? Wouldn’t that be fun to hide a song or do some other stupid thing like that?” And I regret to say that the answer is “NO” because it’s annoying to have a gap of more than a minute on a song… It’s like not being able to finish a song when the key command [End] has already been sent 3 or 4 times in the lyrics and the AI ​​persists and signs by saying “no, well no, I’m not going to stop there! I will continue to generate and consume your credit for nothing until you stop clicking on it and if you ever tell me Get Whole Song I will crash just to see your face of bewilderment cursing because it is a big joke … “. You don’t believe it? Sometimes I suspect the AI ​​of trying to drive people crazy by not listening to the instructions given … I swear that sometimes, it is so frightening it is … that’s it!

That said, if I had to put forward a figure in terms of the percentage of poor workmanship of the tracks generated by the AI, I would say that depending on the period we find between 10 and 20% of waste at the most and that it essentially depends on the styles we use. Anyway, SUNO has already planned: “Less is better” so the less you put in indications, the more freedom you give to the artificial intelligence so that it can make you a beautiful generation… After that, if you come with exotic mixes in terms of sound (like make me some “hardtech poetry rap oumpapa metal symphonic speedcore”) it is clear that you mechanically increase the chances of producing a sound mush in addition to a real cosmic turd… I wouldn’t say that I have tested it but it’s just like… Already mixing “drone, organ, doom gothic, dark” even if it remains in the same vein there are still risks of generating audio vomitive mush (AVM) which are not less and which are clearly more present than when we simply put “Rock Alternative, Dark” or “Heavy Metal, Dark”.

Admit it, you like the dark side too…

At the same time, when the pseudonym of the guy who composes is “THE DARKERS” you’re not going to do Christian Rock specifically… You could, but essentially your raw material is “Dark” and so it’s a bit normal if you find it everywhere (some “Dark Matter”). Anyway, as I said sometimes I have surprises during the generation and I lose one in the process. If it’s the one from the first generation, well it complicates a bit the constitution of the tree made from several pieces but in this case there is always the possibility of generating from the previous branch so it’s not so bad… You just have to not be short of credits otherwise it’s obviously frustrating but otherwise well “shit happens” as they say! As I was saying, it is still quite exceptional to have degenerate generations but for that you really have to push the system to its limits and quite honestly it happens to me from time to time when I want to make a song in a slightly special style but not that much. Artificial intelligence still has a good capacity to work correctly even if sometimes it falls a bit short in terms of lyrics… French is not its preferred language and sometimes it distorts a few words, hence the fact of having chosen generations in English; it still goes down a lot better, especially given the musical styles that do not necessarily lend themselves to lyrics in French…

Anyway, now that I have shown you a little of the dark side of music generation and the errors that can be obtained, I will continue to tell you about my wanderings in terms of my musical project. I will briefly remind you of the context: I discovered SUNO, tested it and saw the potential to recreate a musical repertoire set aside for more than 16 years. The initial project turns out to be titanic and will require hundreds of hours of work, but come on, we’re going to do what we’ve been dreaming of doing for 16 years, and if we’re going to do it, we might as well try to make it a worthwhile project that holds up and that will allow us to free up a few pennies to distribute to a few associations that are close to my heart. It’s also a way to give practical utility to a project that otherwise doesn’t really have one. Why would I compete with real musicians (even if I still am one myself) with my project that uses AI? Yes, people with a moral code still exist. That being said, for the project to take shape, I went through a crowdfunding platform: Ulule. However, despite a plan that isn’t completely out of date, the Ulule team pointed out to me that my campaign has a strong chance of being a resounding failure if I don’t have anyone to support my project. In terms of family and friends, it’s true that I’m a bit light, so I’m going to have to gain popularity… But, the spotlight when you choose “THE DARKERS” as a pseudonym… How can I say it? Make up your own mind!

Well, here we are: I asked Google and ChatGPT and Copilot what they suggested to me to gain popularity and create a unifying community around my project. I got pretty much the same answer everywhere: “Dude, there’s no secret. First, you have to distribute your music as much as possible, and then you have to be present. Make a website where you introduce yourself and your work, release your music, do some teasing, use social networks as much as possible, create your YouTube channel, make a newsletter… In short, be as present as possible almost everywhere you can. Put yourself forward and work on your image. Avoid as much as possible bad buzz and bullshit that will harm you and stop writing articles filled with sentences that contain vulgarities, obscenities or vocabulary words that require a dictionary… In short, be accessible but don’t overdo it either”. Wow… all that? So yes and no… In fact, I developed what my “advisors” summarized in a few points on a “bullet list” but frankly we are not far at all from what they could have told me if they had been human advisors. Afterwards, I tell myself that the checklist is there and that I only have to use it to try to recreate an image for myself… Well yeah because I was popular from 2003 to 2007 when I produced “Les Aventuriers du Château de Morenbach” alone in a context where there were still few MP3 sagas (there must have been 5 or 6 of us at the start) until the explosion of MP3 sagas and their almost total and definitive abandonment with the arrival of YouTube. So, I know what it feels like to be criticized, insulted or trolled live… but also what you feel when people thank you for the work you do and when they tell you that your bullshit made them laugh!

Yes, well, not so much, eh…

In short, the goal here is not to become a superstar but just to convince a few people (a few hundred or even a few thousand anyway… and that’s already a lot to do) of the merits of what I do knowing that tastes and colors… That’s it! On top of that, the project will necessarily divide since you have those who will cry scandal and who will say “He’s cheating, he’s using Artificial Intelligence, it’s not real music, it’s fake! He’s trying to trick us!” and the others who will say “Well, after all, why not? The guy is using the latest technologies to revive his own discography and share what he’s done, what’s the problem with that?” So we agree on the principle: we will certainly have some people who will literally shit on the project as it is presented for various reasons… It remains to be seen whether they constitute a majority or a minority of people in the public… And I will only know that when I have launched the first singles… Because yes, it is already on the way!

So it is on the way, but it will not happen right away… I scheduled the first release for 4 weeks and I crossed my fingers. But how did I get there? Well, I started by rereading the TuneCore conditions a little and I said to myself “Ok, I can easily pay €39 per year for a semi-pro subscription since I will have a 3-year period to release all my albums at a regular frequency. Incidentally, I can already subscribe to the TuneCore Publishing option even if I can’t use it at the moment…”. Oh yes, but I have to explain because it’s worth its weight in peanuts! So in terms of law, SUNO recommends that you protect yourself and your music. Did I say that SACEM wasn’t keen on AI? It’s still relevant, but but but, I received a response that changes the data of the problem a little and that I’m going to share with you here (just the response).

Finally, SACEM is perhaps not as rigid and closed as I thought… Do they need money?

So, when I read that, I really want to go get SACEMized but it’s possible that I’ll get SACEM-sodomized because it will depend on the appreciation of the members opposite and pay 100 bucks to be told “uh yeah, bro… you’re nice but when you just write the lyrics of the music your personal contribution is nothing!”. But what? They protect the lyrics well so… it could pass (I wrote a text wrapped in music generated by AI. If we consider it as lyrics, does it pass or can I make people believe that it’s a song?). Okay, I admit, I’m walking on eggshells and this is perhaps not the time to be ironic, cynical or ironic… nevertheless, with AI it’s always a bit of a sticky and slippery subject that seems difficult to put on the table given the legislative void, the controversies and the general opinion of musicians who tell you “no but AI doesn’t replace humans” when in fact that’s not what we’re asking of it: we’re just asking it to serve as an interface to generate music that we can’t or can’t make anymore in a group from what we’ve already prepared ourselves. In any case, that’s valid for me… But to get back to SACEM, I want to say that in addition to the 100 bucks it’s going to cost, there are also the membership conditions that are a bit annoying:

The conditions necessary to justify membership.

Not so long ago, the conditions were a bit more complex than these but already now it’s going to be complicated in the immediate future. Explanation: A capture of a YouTube channel with at least a thousand views well… you’re going to have to create the channel, feed it and popularize it and quite honestly with my family life and my job it’s going to be complicated to set aside time to make videos when most YouTubers have almost nothing else to do all day long since they live on the monetization of their channel and therefore the content they produce. So be careful, I’m not saying that they jerk off their balls all day long… far from it! In fact, they’re going to spend hours taking care of the editing, the takes, mixing… A bit like how you make a short film and honestly, it’s consuming in terms of hours. Even to make a video of 1 minute max in Tik Tok format can take a crazy amount of time! So me who already writes my articles at night during my sleeping hours, if I have to stuff myself with a YouTube channel it’s dead… And yet, I’m going to have to do it anyway but I’ll come back to it…

A double-sided copy of the CD, Vinyl or DVD cover. It would be super easy to do since I’m already doing my own artwork for the crowdfunding campaign on Ulule… Well, graphic design isn’t my cup of tea: I know how to do it but I’m not very good at mastering Photoshop or other tools… For the moment, Copilot is in charge of generating the horrors that I ask it to create and it’s going pretty well for making covers… Now, making a complete booklet isn’t that easy and so I might hire a graphic designer or someone who tinkers a bit more than me in the field to come up with something that’s a bit classier than the horrible rag I’m preparing… But if all I had to do was throw in the cover, that would be very easy, except that what I’m also asked for (and it’s not written on it) is the barcode generated during the production of the work with a view to its marketing. And there, obviously, it will immediately be more complicated given that I will first do digital on the platforms to make myself known, create a community (or try in any case) and then launch the crowdfunding project that will allow me to do physical format. So it’s not for right away…

The broadcast certificate, for the moment it’s “not even in my dreams” because I would be surprised if my works started to be broadcast on TV and radio… As for doing concerts and shows, I think we can forget about it! So we’re going to fall back on the last possibility: Broadcasting on a streaming platform… And there, we’re going to have to start praying or even incantations because my first track “Truth” comes out on September 30 (the time for the platforms to integrate my profile and my information) and if it’s on TuneCore, they guarantee me distribution to the platforms but not the fact that the platforms agree to broadcast my track or not (in fact, I sent my tracks in a group of 3 like an EP). Afterwards, I packaged it as best I could and followed all the instructions but you never know. So let’s summarize: I took out a semi-pro TuneCore subscription at €39/year for unlimited releases and I subscribed to the TuneCore Publishing service to recover all my rights but for the TuneCore Publishing service to be active, I need a SACEM ID number (so I’ll spend 100 bucks but it’s once in a lifetime, it’s okay) that I will only have if my tracks are broadcast on a streaming platform and I can provide proof. Who said I like to live dangerously? In fact, everything is conditioned and I gave myself time to be able to carry out all the other operations that I will have to carry out afterwards… You’ll see, there is plenty to do! It’s not easy to create and distribute your own music when you know nothing or almost nothing about it…