Composing Music with Generative AI | Misici | Skillshare

Playback Speed


1.0x


  • 0.5x
  • 0.75x
  • 1x (Normal)
  • 1.25x
  • 1.5x
  • 1.75x
  • 2x

Composing Music with Generative AI

teacher avatar Misici, Music Composer & Producer

Watch this class and thousands more

Get unlimited access to every class
Taught by industry leaders & working professionals
Topics include illustration, design, photography, and more

Watch this class and thousands more

Get unlimited access to every class
Taught by industry leaders & working professionals
Topics include illustration, design, photography, and more

Lessons in This Class

    • 1.

      Intro

      2:45

    • 2.

      Getting Started

      3:07

    • 3.

      Generating AI

      12:14

    • 4.

      Enhancing the AI

      13:17

    • 5.

      Onward Toward the Future

      2:26

  • --
  • Beginner level
  • Intermediate level
  • Advanced level
  • All levels

Community Generated

The level is determined by a majority opinion of students who have reviewed this class. The teacher's recommendation is shown until at least 5 student responses are collected.

85

Students

2

Projects

About This Class

For those with a curiosity toward the future of music, composing music with generative AI has arrived at last. If, like me, you've been watching artificial pop stars take the billboards and stadiums across the US and Japan by storm, you must be curious about the entire process. 

In today's class we'll be using the most cutting edge generative music platform to create our own track, then we'll clean the track up as best we can in Ableton Live. It's a relatively simple process that can lead to startling results. 

To get started all you'll need is a laptop with a web browser as well as a DAW such as Ableton Live. I'll also show you a few plugins you can invest in if you'd like to take your AI tracks to the next level. This class is well suited to students at any level who are interested in taking that next step toward the future. 

Meet Your Teacher

Teacher Profile Image

Misici

Music Composer & Producer

Teacher

Hi! I'm Jordan, I also go by Misici.

I'm a composer & producer from Australia, having studied music at the University of New England. I also teach music and performing arts from my base in Shanghai as well as in partnership with institutions in other cities on request.

I score projects on a freelance basis for animation, video games, and film. I also produce lofi and dance tracks for Spotify and YouTuber clients. I fell in love with music production while rocking out to the Doctor Who soundtrack in my car with my best friend as a teen. Since then I've been obsessed with finding and creating the perfect leitmotif.

My favourite style of class is short, sharp, and focused on creating and refining a single track that's applicable for a focused purpose. I don'... See full profile

Level: All Levels

Class Ratings

Expectations Met?
    Exceeded!
  • 0%
  • Yes
  • 0%
  • Somewhat
  • 0%
  • Not really
  • 0%

Why Join Skillshare?

Take award-winning Skillshare Original Classes

Each class has short lessons, hands-on projects

Your membership supports Skillshare teachers

Learn From Anywhere

Take classes on the go with the Skillshare app. Stream or download to watch on the plane, the subway, or wherever you learn best.

Transcripts

1. Intro: Hi, Are you curious about generative AI, especially when it comes to music? Well, if you are, you're at the right place. Hi, I'm Jordan. And I am so curious and excited by generative AI and especially its application when it comes to music. For the past couple of years, generative AI has been a huge topic of conversation all over the world. Especially because of chat PT and everything it's been able to do for us in the world of writing and pictures and all kinds of things. And it has led to the development of a lot of technologies. Well, music is no exception. Ai artists have been blowing up the world. Especially in countries like Japan and the US. Where AI artists have overtaken Spotify are selling our concerts and arenas. It's been unbelievable the progression of generative AI over the past couple of years. For us normal people, it's been a bit difficult to get involved. The process by which we've been able to make AI music has been complicated. Not so easy, and the quality has been just awful. But lately it has gotten much easier for us to get involved. The quality has improved a lot and the methods by which we make the music have gotten a lot easier. So if you're curious about generative AI music, let's get started today in making our own, we are going to make our own music and then we're going to clean it up using some conventional methods and plug ins that I would use for any other music. We're going to make it on the platform that I trust the most as it stands now in January of 2024. Of course, technology is always changing and updating and improving, so we're going to have to keep our fingers on the pulse and update as the technology does. But the music that AI has been able to make has been so exciting. I mean, just listen to what I've been able to make with AI just recently. Consationuch. I mean it speaks for itself. I mean, who doesn't want to make music like that and if you don't have sings at your disposal but AI can sing for you, I mean, why not? So if that sounds interesting for you, if you feel like you're someone that wants to make AI music and wants to get involved, I mean, join me. Let's make some AI music together. Let's get ourselves some awesome Beats moment. We've got a phone. 2. Getting Started: For the music we're going to be making today, we are going to use Sono, that's N. I have found that that has been the most reliable source for AI music right now in January of 2024. Previously you had to make their music on discord. You had to go into their discord and you had to make the music that way. And it was, it was a trial, but now we've got the web version and it's awesome. No app yet, but web version is just fine. You want to type this into the web browser, press enter, get into their website, you're going to be met with a splash page. And you just click on the top right, make a song and it's going to load the apps web based app. Now once you're in, well in my case, you're going to see all the songs you've previously made as well as Explore, create, and library Library has got everything you've made before. Creates your way of creating and to find what other people have made previously. Now down the bottom, toward the left, you're going to see my credits. Credits is an interesting system. Basically, you have enough credits currently with the current system that you can generate five times per day. This is enough to make a song each day as long as you're not too picky about what's in the or how the song sounds. Because every time you generate, it's going to automatically make you a melody in a baseline and progression a sound To get a new one you have to regenerate again. And each time it uses credits, you have enough credits to do that five times per day for free. This is enough for a lot of people. That's fine. But if you're like me and you want to keep generating and you want to keep playing with it, it's not enough. You can get a subscription. Here you can see the prices. It's right now, $8 a month for the pro and then $24 a month for the premier. I can't imagine anyone really needing the premier. I don't know why you would need that much, but the pro level is pretty awesome. Okay, once we are going to create the songs, we're going to do so by clicking Create. Making an account is pretty easy to logged in. I just made an account with my e mail. It's really funny, when I made my account, it had the option to sign up with Google. I did. I click sign up with Google. On Google's end, it said, This website is not verified with us, this isn't going to work And then it kicked me out, so I just sign up through my e mail, the manual way. All right, now we're in create. All right. Sweet. Let's make some songs. 3. Generating AI: Now there's a couple of ways we can do this. Make a song, this information, this is how it is in January of 2024. But we'll need updating sooner or later because it is progressing all the time. You can watch guides from last month that are outdated. There's a couple of ways and the first way is going to be the simplest, which is to use chat chat GPT to give you some lyrics. I structured my chat GBT lyrics already using these bracket types to tell so how I want the song to be structured. I've noticed that as of now it recognizes intro verse and chorus, but that's it. I'm not having any luck with anything else such as bridge or Altro for now, at least. Intro verse and chorus is what you're going to have in this part where I have a separate bridge and chorus. I've just got it here as two choruses. Yes. I asked chat GPT to make me a song about skillshare because why not? That's what this is for, right? I'm going to copy this and paste it into here. We're in Sono now we're going to go to custom mode, put the lyrics in just like this style of music. I'm going to say it is female. I generally like the female AI voice better female EDM dance track, High energy, Happy, Create. It'll always give us two options. Let's see what we got. I should have mentioned that you can put a title in. The title will have an influence on the art that it generates for you, but I didn't do that. Anyway, it'll take a couple of minutes or maybe 1 minute, and it will make the song, when you've put in so many lyrics longer, it'll take, let's see, the first one, no guides us shining through the light. That's okay. Let's say the second one given knowledge, we'll reaching new heights. Skill show guides us shining through the night. Thank you again. It's okay. But judging by the length, I can see that I don't think it has made everything. I asked for intro verse, chorus, verse, chorus, chorus. It says it has the winning beginning knowledge to be in new heights skin experience for breaking the hole. No, it says it did, but it did. I guess right now we're at a minute. 20 is the max that I can do, which is a good. There is another way that we can create. We could use either one of those. There is another way, which is we're going to start with just letting it go crazy. We're going to write in create a high energy EDM dance track about learning on skillshare Create. Then let's see what it does. This is just the general non custom mode. It can just go wild with this mode. Oh my God, skill share high. I'm already obsessed with this mode. It's going to be only about 30 seconds long, and then we keep the track going a few 30. The must be found, just for the only thing I don't like about this type of creation is that it seems to start halfway through something. See, I like it, but it doesn't start at the beginning. It's not very useful. Yeah, a verse and a chorus. Otherwise it's good. I love the name. Oh my gosh. Let's try. To do the same thing again, but just add an intro. I'll keep the intro that's already there. Put the verse and chorus in that it just came up with. And then that's it. See if we can continue that, taking away the word happy. See if that leads to a better result. Maybe I'll add the word trap. Let's see if that makes something a bit more interesting. I forgot to re, use the name. Okay, we've got our new tracks. Let's see. To heights, shine the region start to punk the knowledge. Still there's a world of learning just for guides shining through knowledge. Heights shine the region, start to unlock the knowledge still. So there's a world of learning just for you. Still shares to. All right. I'm happy with that. We're happy. We've got 43 seconds of the beginning of the song. Let's keep it going. Click these little dots We're going to continue from this clip, takes us over to here. To continue it, we're going to need to generate our own lyrics. So I'm going to delete what we have. Maybe I'll keep the chorus going to copy that. Then delete the rest and click, make random lyrics. See what it comes up with, Roll those dice. Okay, we've got a verse, we've got a chorus. Now I'm going to treat that like a bridge and copy back in the original chorus. Continue. Now see it's coming up with two options for part two, which is good. It's going to keep that song going. Now we get to pick which part two we want to use. This is actually my favorite way to do it. It's a bit more random then copying in an entire list of lyrics generated by Chat GPT at the beginning. That's a favorite of people that want to feel a bit more sure about how it's going to turn out, they'll copy in those lyrics and then they'll click Create 25 times until it comes up with a beat that they're happy with. That is a good and dependable way of doing it, but I like this way each part is separate. It thinks about each parts differently, although the song is going to be really wild and unpredictable with each part being quite different melodically, although that's not always such a bad thing, you know. All right, it's done. Let's see what we go. Walking through the knee on city lights where the stars don't shine dreams at night. It's quite similar in style more than I was expecting. Walking through the knee on city lights with stars don't shine dreams at night. Oh yeah. Everyone's hustling, trying to make the way in this concrete jungle. Music I'm happy with that one. Click the dots again. Let's do it one more time. Continue from this, let's delete and one more time, make random lyrics. It is using Chat GPT technology on our behalf. Yeah, and chorus. I'm just going to put in our old chorus to finish up the song and then that'll be it. I'm going to try again to make an outro and just see if it works out to should I have one tea or two teas? I tried it with one tea, it didn't work. Should just be one tea. I'm just going to make it one tea. Okay, that would be really crazy. Okay, let's do it part three and then I reckon that's good. Three parts, although you can just keep going as long as you want. It should be mentioned on the free version, you get 50 points a day or 50 credits a day. Ten credits per clip per generation. So you can do this five times five clicking of the create or Continue button per day. I obviously use those up so quick. I paid for the $10 a month subscription, which gets you 2.5 thousand credits. Still at ten credits per generation. About 500 generations. I believe that's what I did. Those credits will last me the month, this not per days, per month. 350 per day paid. 2,500 per month for someone like me that wants to go mad in the city streets on dreams. Bright lights around those blinding beams. City streets on dreams, lights around blinding beams. All right. I'll like that second one. Okay, we're done. I'm going to click the three dots and click Get Song. It's going to put our three selects all together into one complete track. And we can download the AI generated video it's made for us, or the song or both. I'm going to download both. Then from there it's done. Then from there we can take it into Ableton and do any mastering or quality fixes that we want to do, whatever we feel like. 4. Enhancing the AI: All right, well here it is. I've dragged it into Ableton and it's already so impressive. I can't believe how good it is. So what I'm going to do is I'm just going to add some stuff to kind of give it some extra punch and a bit of extra clarity. Because as of how things are right now, you do sacrifice a lot of quality and a lot of depth in these AI tracks. First, I'm going to supplement the kick drum with just a general kick clip that I have. I'm going to just drag it in, try to line it up, and then just use it for the chorus. The chorus has quite a strong kick. I want to supplement it. All right? I only have to line it up once. I can just drag it across to the other chorus parts there. Yeah, I think the quality of how it exports is what's keeping regular people, composers, safe for the moment. I think companies can get high quality bounces from the AI every day. People can't yet. It'll only exports an MP three. For now, it's just a low quality sound to it that is temporary. Look at our fastest technologies are calm. It wasn't so it was very recent that it was even invented. Even more recent that it's become available for everyday people to play around with. All right, I'm going to put ozone over it and then just protect our ears from being blown out. I'll just do it with ozones. Automatic assistant. Master assistant man, I love this plug in. It's not free but it's highly recommended. It just listens to your track and has it so that things like this get fixed. Okay, cool. Isn't that just such a beautiful just sound like it's just wild? Um, yeah. It's not going to be long until it doesn't need fixes like this. Okay. What else can I do? Oh God, because the vocals are pulled into it. I can't do my usual stuff to improve the vocals. I can see if it will improve the vocals if I isolate them and then try to work on them a little bit, but I don't know how much it's going to help. First, I'm going to use clarity. I'm a big fan of clarity. It's just going to isolate the vocals. I'm just going to take, I'm just going to meet the first one. Knowledge still speak. Once the kick comes in. Man, it just, it works so hard it can't find the vocals anymore. We lee, I don't know if this is kind of like what I usually do. I would also, I would also put a Vd over the vocals as well, but with them. Meshed in with the instruments. It's hardly worth it. Let's put it all together, all right. I do think it helped a little bit and I think that ozone has helped a bit as well. Overall messing with it in Ableton hasn't done nothing. It has done a little bit. It's just so funny. All right? Yeah. So this will be our track. I'll just play it through one time so you can hear what the whole thing sounds like we were given by the AI gods in total. Quick, I figured something out. I'm going to go back to our song here. I figured out that while it doesn't stand out, it does understand intro. I tried this with another song first. Let's see if it works with our song. Let's put another intro onto the end of the song and see if it acts like an out. We're going to continue this clip once again, but this time from the full song. We're going to make random lyrics, see what it gives us, but it doesn't really matter what it gives us. Because we're going to cut in our existing chorus, or I need the word chorus. We're going to put in our existing chorus, there's a verse there. We're going to cut it short. Call it an intro. Put it on the end. Let's see, Now I tried this already, it still didn't give me a very satisfying ending. I'm finding that no matter what I do, I can't get a real ending on any of these songs. The last time, the last one, it just ended. Let's see what we can do about that, if anything. All right, and continue from the top. I wake up the sun shining to bring you. But I feel something, I can work with that. Feel something in the air. Oh my God, that's so creepy. Alright. I can work with the first one, so once again get whole song. Let's copy that into Ableton. All right? I've copied it in, you can see it's much longer than the other one was. Let's make sure that the warp is turned off. There we go. But it should still line up with all of the things that we've already done came. Oh my God, Chop that off. All right. I'm literally just going to cut it and then I'm going to make a new onto which I'm going to add the last part. I've still got that weird, creepy bit. All right, chop and paste down here. And then on the new track I'm going to put a reverb and then just reverb it out. Perhaps. Swelling synth, Let's see, Not too bad. Just too loud. All right. Better than nothing. A week ago, the sun turning to bring. It's not perfect but it's song length. It's doable. It's definitely doable. All right. All right, that's good. Let's, let's call it here. I'll put on a full version that you can watch if you'd like to. Knowledge the rhythm. Knowledge of learning just for you still shares some that is not. Camels never stop give walking through the knee on city lights with stars, on shadow trains tonight. Oh yeah. Everyone's hustling, trying to make the way in this concrete jungle where the music plays, lights got through the night, colors, sights in the world. Never speaks the form of life. Snoopy were skin. Next level level, hands on deck, ready to care levels, never start. City streets, which is non dreams. Bright lights around us blinds. And beams were lost in the crowd, but it feels like home dancing to the rhythm. We're never alone. The next level, hands will never stand a in the top in which is star we can met levels high. We were reaching the top a week up. The sun is shining a brand new day. 5. Onward Toward the Future: And that is it. We did it. We made music using generative AI. Thank you so much for joining me today. It's been a real pleasure to experiment with Sono with you and to fix up our beats and add that extra kick and, you know, get skusher high made together. I have no idea what I could ever do with this track. I don't know if it has any practical use at all, but it was fun to make and it's fun to experiment. If you have the $8 a month to play with. I recommend getting the pro plan and just keeping on generating tracks and experimenting both with writing your own lyrics and with using the lyrics that the AI comes up with. I found that by putting it a prompt and having it write lyrics, I've come up with some real bangers and I've come up with some more Duds than bangers. To be honest, it's fun to play with experiment to get more used to the technology. The important thing to remember is that this technology is the future. This is the future of composing. In my opinion, it's quite a controversial topic. People don't want to think of creativity in this way. They don't want to think of the idea that the composing process and the creative process is under threat by AI. But it is, I think that the only way to future proof yourself is to become familiar with this technology and become someone that can manage it and grapple with it and can run it that way. You're protecting yourself for whatever the future holds. What if one day the future of composing is managing AI is arranging what the AI has come up with. Choosing between one melody and the next every 30 seconds and stitching it all together. As I said already, quality is holding this AI back. But for how long? Look how much we've gotten in such a short amount of time. How long is that going to be true for? So keep playing with the II. Keep experimenting and do let me know if you have any questions or you want anything else covered at a future class. Thank you so much for joining me here on Skillshare. Please look at some of my other classes. I've got composing, I've got music theory, got a few things. So thank you for coming and I hope to see you again soon.