That face when I've been prompting manual mode style without turning manual mode on. THANK YOU for that, that makes so much more sense why I wasn't getting the same results that others were.
Be careful though, "-1" is random and you should not mess with it. If you roll the "dice" it will keep the seed that comes up for that whole session and you will get no variation. The only time you should mess with the seed is if you have a song that you want to tweak a little but want it to basically stay the same, at that point you lock that seed in and new iterations should be very similar.
Oooh, this video was recorded while my song ("Paid For") was on Staff Picks. That's really cool! Great tutorial. Glad to see in-painting in a video format. It took me forever to figure it out, for some reason.
Great video, I have been working with UDIO for a long time and this video has taught me so much. I'm going back to work on a couple project now that I have way more control Thank You Keep Creating Your changing the world !
Ahhh, thank you so much!! And to be honest: Thank you to Udio for sponsoring the video-- It's because of that, I could spend more time really digging into the various features. Plus, I had them as a backup to double check and make sure I was getting everything right. Hoping to do more of these deep dives with them. Now, in the meantime, the next 4 videos will be sponsored by energy drinks and Skillshare. HA, kidding!!!
I learned the hard way you can't really write too much in the 30 sec version as it just gets cut off, but I did get some dope results from the 2 min version. Did you know you could tell it to things like stand up comedy by using [dialogue] and prompting for a stand up comedy routine. Works better if you feed it the jokes. I was also able to have it perform songs in multiple languages in one song. It's fun to try to see what other things aside from just music you can actually get udio to reproduce.
I have an issue where I like 2 extensions of a song, but everytime I try adding another verse, it'll sing to me gibberish, or if I try to replace it with an instrumental, it'll give me the pre-chorus lyrics I inserted previously. It's annoying. I really like the first 1 minute and 30 seconds of the song I created. I want it to be a 5 minute song.
14:15 nice playing! sounds really good. My favorite track here was 7:54 :) Udio and Suno are incredibly amazing tools of the modern day. I got to extend a track my friend made and we never felt like it was finished, so much fun.
Appreciate that! I’m rusty in the woodshed (the channel keeps me pretty busy), so that’s super great to hear! And yeah, that generated song at 7 mins was great! Love where we are with AI music, and it’s only getting better!
Wish people would not difference between typing a few words and letting the Song AP completely develop a song from a song writer using one's own lyrics and using the Song AP to work the melody and refine your lyrics. When you say you are using AI song AP everyone thinks you just type a few words to get a song. Many do this, but song writers do not. We use our own lyrics.
Agreed. I mean, I totally cheated for lyrics in this video using ChatGPT for the "written" lyrics-- but, that's for a tutorial. If I were crafting a "real" song, I'd write my own. That said, I do think the various AI tools are great for springboarding ideas. Sometimes you can get stuck, or just need that little spark-- and AI is great for that. More often than not, it'll provide me with something I DON'T want-- but that is enough to get me going. y'know?
Udio has superior voice quality and more realistic all around audio quality. However, after the last update, it seems to produce a lot more gibberish and odd vocal artifacts (on standard settings). It became almost unusable as each generation would be ruined by undesired and weird vocals (even on instrumental setting). It's a shame because I loved Udio despite its other quirks. The latest update seems to have broke it. Had to cancel unfortunately in favor of the reliability and consistency of Suno. Thanks for the great vid!
@14cassara I also had a similar experience with UDIO. I absolutely loved it but feel like the generations are getting worse. Suno is much more consistent with getting decent sounding music, I recommend Suno for now.
Yup! I didn't get a chance to dig into that part, but-- we're talking about doing a follow up for the "power users" crowd, and that will for SURE be going in!
This was helpful as to inpainting, which I've not tried. You didn't say, or I missed it; as to whether using the same seed to continue a much longer song will keep it sounding similar throughout.
hypothetically, it should. Although-- y'know, it's still AI so anything goes. But, weirdly-- you now have me wanting to experiment with changing the seed on an extend to see what will happen. That might be interesting...or a train-wreak. ...probably a train-wreak, but maybe an interesting one!
Please teach us - Clip start and lyric timing in conjunction with remixing and inpainting, or just in general with real examples. It's confounding but seems important!
The ability to upload a melody line changes the whole world of AI generation - that was just fabulous. Is that a paid only gig? (worth it, I might add)
Great overview and guide. It'd be helpful next time to mention what requires a Paid Subscription; Udio is quite generous with their Free-tier, however: Inpainting, Stem Downloads, and Custom Cover Art requires a Paid-tier.
Ah, that’s a good call. I think I had that in an early outline of the tutorial, but it might have gotten overwritten as so many other sections developed out. I was aiming to keep this around 20 minutes, and there was a LOT to cover! Great point though!
Hey, was beginning to have Tim withdrawal symptoms. 😆 Been playing with it for a couple of months now. 1. 17:13 What settings did you use to get the inpainting to _actually change_ the words? Because my goodness, it's been a slot machine so far. 2. The clarity slider will hopefully improve. It's fantastic that we can download stems, but a little challenging when there's bleed from one to the other. If the slider is tweaked too much, it starts sounding very unnatural. Basically stem separation at the cost of synthetic sound. 3. 5:19 Thank you for these. I didn't know about [Refrain] 4. For the love of Thanos, please Udio, give us negative prompting for words like say, echoes, pulsing, shadows, pulsating, neon, city, etc. 5. Using stems from different iterations and editing/mastering them in post is awesome! 💯 They absolutely need to work on the acoustic/organic instruments' sounds in v1.5. They very frequently sound like synthesizers as opposed to v1. eg, violin, cello, flute, etc. What? NIN uses flutes! And the whole 'loudness wars' generation approach, pre-baked compression and bell-curved eq'ing doesn't help in the mastering process. Depends on the use case I suppose. Fantastic video Tim. Even a short follow-up video about what you find with the settings would be greatly appreciated. 👍 Sorry about the long comment, but hey, it's generative music!
Oh no need to apologize for the long comment! Like I said, I think Udio will be looking into these comments, so it's great to have focused and realistic feedback! For the Inpainting, I'm not sure-- beyond the *** open/close, I did also write lyrics (well, chatGPT did) where I instructed it to keep the same meter as the original. I don't know if that plays a part. Like, if you try to inpaint out a club banger with like, lyrics inspired by Tales from Topographic Oceans-- it might not catch. hmmm...now I want to go try that. NEXT STOP: Prog Concept Album Dance Bangers! haha
@@TheoreticallyMedia This is the wrong way to inpaint. Read the instructions carefully. You need to select the section you want to change, but when you tag the lyrics with ***, it needs to encompass all the lyrics that are present in your 28 second window. Once you do that, no more gibberish
Hi, great video I find Udio clearer on voices than Suno, however, out of the blue, which tool would you use for a music video? Kaiber? Noisee? Another one? I'd be interested in having your thoughts about that. Thanks.
I had never thought about changing the lyrics through in painting. Maybe this is working better now in 1.5 but I recall in the early days that I often had very wonky results when inpainting areas with lyrics.
I mean, it's still AI-- so you might get some wonk-- but it has gotten a lot better. The other big trick a lot of times is to spell phonetically. And it helps if your inpainted phrases match the meter of what you're tying to inpaint. At least, that's what I've noticed in my experimentation.
That’s a great thing about a lot of the AI models. Midjourney included: the fact that they don’t bury the old models and you can still access them. It’s all iterative, and to get to 2.0, Udio has to do 1.5. Some things will work better than others, some things will break, but it’s all part of the road. Still, props to any platform that lets you continue using the “old” model. Haha, old, it’s like 4 months ago!
I think it would be cool to have a slider for how "normal" or "crazy" you want the output to be. Sometimes it goes to far in one way or the other. A slider for how repetitive or chaotic you want it to be would be useful too, as well as one to decrease and increase intensity for a section. Better support for specific instrument tags like oboe, xylophone, slap bass etc. would be awesome.
outstanding video bro, I love Udio and I learned some great stuff from you just now thanks a million. (I may have some sweet instrumentals out there too) Thanks Tim!!
I want to see, uploading parts of my songs (which I'm not the best at performing or singing) and getting Udio to Upscale it into a high-quality version. I'm curious about how well we could nail that. Like upload the verse and then extend, have [Verse Repeat] and the exact same lyrics. Then do 20 versions to get an almost exact copy but where the singing is perfectly in key and the guitar is super tight without buzzes and the mastering makes it all sound like a studio production.
A few videos ago I did something similar, where I recorded kind of a blues-y white stripes type riff, that I thought sounded pretty good-- and then Udio came back on an extend with mostly the same riff, but man-- that timefeel was SO much better than mine! Haha, I ended up relearning the Udio version! I'd say keep working on your own talents and use Udio as well! And, if you bone up on production, you can always take the Udio stems (which I didn't go deep on here) and bring them into a DAW as well!
Thanks, great video! We have used Udio to make some fun tracks. Also have seem some people doing amazing cover versions. We have tried as well, but without any success. I presume that it's only possible with public domain lyrics, but could not create any with PD lyrics either. We tried prompts like 'cover of ........' etc. Did Udio change something in 1.5?
Some very helpful explanations. How do I cut out a section? Also, there does not appear to be any way to control the loudness of a section that is way too quiet (obviously not the playback volume which affects the entire song). Thanks.
Obviously not Udio, I’ll pass along to them for clarification. Offhand, I might try [applause, tires skeeching] (I don’t know why that particular example came to mind, but man that’s a weird scene) Or, [tires screeching] and then [applause] on a second line. Udio tends to be pretty smart and will likely figure out that you want those two sounds blended. If you want SUPER fine tune control? That’s when you might be moving into generating two independent sounds and working on them in an audio editor
BTW, it's so weird because I could have sworn that in the comment section of your last video I congratulated you on your 100K Sub achievement, but I cannot find that comment. 🤔 So in the event that I'd wanted to congratulate you but for some reason got distracted or something, I'm congratulating your milestone here! 🥳 🎉 🪅 *CONGRATULATIONS!* 🕺 💃 🎂 👏 👏 👏
Thanks so much, Louis!! Yeah, comments go vanishing all the time these days-- I don't know what's up with that. But, man-- thank you so much. You're a big voice here, and I always appreciate reading a comment from you!
Two Questions: If you are creating a song in 30 second intervals and you arrive at a section that you want to be instrumentals how does that work? How do you get the chorus to be the same melody from one ext to another or should you just go to the longer time limit?
What I want to know is if we remix our own audio (like a guitar recording) or even just make one using a prompt and record something ourselves that sounds the same, we're aloud to use it ourselves right?
1000%. If you're talking about from a copyright standpoint, I think the current legality is that you can, provided that you can prove that there was a "significant" human element to it. What does that mean? No one knows, and no one wants to challenge that-- since that means courts can define what art is. Basically, yeah-- you're fine.
Udio can only remember 2 minutes at a time, so if you create a 15-minute song, the end and the start can be completely different. Also, we have the option to create 2 minutes of the song at the beginning; once again, if you create a 15-minute song, resorting to 30-second increments after that can take you all night. We need the option of continuing with 2 minutes at a time. I create a lot of Rock Opera type songs and have found it can take a couple of days just for one song. (I'm very picky, so out of a dozen iterations, I might get one I like) Hence, I've returned to a basic plan because I don't get to use even half of my credits. All that said, I absolutely love Udio.
Yeah, you’re hitting the nail on the head of something I preach quite a bit on the channel: AI is not “push a button” magic. It’s a tool, and if you want to make something great, it’s still going to take work. Less work than trying to arrange 4 band members to try and get in a room with a recording engineer? Yes. Less cost than hiring a string quartet to do that conceptual bridge section you want? Yes. Less time than trying to sober the drummer up for that fill that he said he’d punch in this afternoon? For sure. Speaking of which, we really need to fire that guy, he’s just a mess. Totally get where you’re coming from though. And as a prog/concept album guy myself (Yes, Rush, early Genesis etc), I TOTALLY get “picky” The same goes on the AI video front. As a filmmaker, it’s a challenge to “direct” AI. But the fact that, after some elbow grease and time, you can realize some amazing piece of music or video? That’s kind of amazing. All that said, it looks like I’m preaching to the (ahem) choir on this one! Keep it up and always remember: this is the worst it will ever be. It’s going to get better, faster, and smarter. But, it’ll still always take a guiding hand. And, if you’re wired like we are, it’s still always going to take work. Ps- I don’t know if you’re a Zappa fan (I’m not a hardcore one, kind of mid-ground), but I often wonder what he’d make of AI music. Obviously (if you like his music or not) he was a genius- and also quick to jump on new technology. I tend to think he’d be simultaneously fascinated and frustrated by AI. Probably would have hired some developers to make his own model and drive them all insane with 2000 daily requests for features!
@@TheoreticallyMedia I am a musician , and I ve been using UDIO to a great extend. I love it. I have the basic plan and , honestly, I use like 500 credits to get what I want. I also upload A LOT of my stuff to be based on. With that being said, I use Reaper and Kontakt afterwards, Editing, adding, cutting things. Honestly? It feels pretty close to composing a song in my guitar. I ve always had problems finding the right vocalists. So...I had many songs "stowed". Now, I can at least release them (well..I did on my chanel ) 🙂
If you do a deeper dive, perhaps you could give us some examples of how the various sliders impact a song rather than just telling us. Plus a bit more on crop and extend. Also Remix of the first 32 second variations can be interesting. Crank it up to 100% to get some significant variations and then extend your favourite
I was hoping for more info around working with the different stems of a song? Is this a thing in Udio? Or do you need other software (which ones) after you download stems? Thanks in advance!
We’re going to do a “Power users” video in a bit. I know, it’s a weird balance doing these things. The general audience that mostly watch YT are looking for “quick” answers- but like you (and a lot of commenters here) we want the under the hood stuff. The idea was to start with a basics to intermediate video, which I think we nailed, but the next one will be more focused on the Pro side.
Welcome back, Tim! I was going through AI withdrawal... I'm guessing it cannot, but I'm wondering whether Udio can do classical music, anything from Gregorian chants up through 12-tone stuff, or even just classical or romantic.
Best bet here: Have a good chord progression/riff, run an extend with a prompt in your genre-- then download the drum stems and pop them back into your composition. Seems to work pretty well! And, you can also download a bass player! haha
I use inpainting a lot on the instrumental parts of songs. I haven't heard anyone speak on the creative things you can do with inpainting to change wonky notes in a solo section, alter a drum fill or other things like that. I've figured a lot of things myself, but since this is all so new, every new piece of information is potentially helpful.
Suno does better with uploading audio and is super fast compared to Udio. But sound quality wise - Udio is incredible. Also you are the guru of AI - love your channel and content.
Ahhh- Thank you so much!! I actually haven't played around much with Suno's upload feature-- but what I like about Udio's is the idea of remixing a barebones track to use as filler around it. This video was obviously running super long as is, but I'd love to do another one focusing in on the "power user" aspects. I think there's a lot to discover here! Haha, I will say: We've gotten so spoiled with speed lately! Like, I've got to wait 1 minute? Why isn't it done in 30 seconds?!!! Just makes me laugh, since-- yeah, I totally do that too. Meanwhile, 2 years ago we would have happily waited 15 minutes for something like this!
I got a subscription for Udio and also use Suno sometimes. I gotta say that these 2 a quite different and I’m happy to use them both. Udio is more specific - you can upload your own audio and Udio might add vocals and just some subtle things to the audio you provided. Whilst Suno usually changes your input much more. And I wouldn’t consider it as a drawback, because quite often it comes up with some unexpected and helpful results. As to audio quality, I totally agree with you - Udio is much better in this regard
@@TheoreticallyMedia Haha yeah. I've spoken to the leadership of Suno and they have great things in the pipeline and I told them that sound quality needs to match Udio really...and talking of speed - I love RunwayML Gen3 turbo feature for Image to Video which is AMAZING - fast and prompt accurate (for me anyway). I'm getting stunning results compared to the non-turbo version.
Ah, that’s great to hear! I feel super rusty, since the channel keeps me so busy and I don’t have as much time in the woodshed! Appreciate the comment! Still, gotta find a little extra time in the day to touch wood! Ahem…wait, did that come out wrong?
What happened to Runway Gen 3 lowering their prices? And don't say Gen 3 Turbo is the new "cheaper" option because it's lower quality too, so that doesn't count, hehe!
You know what, you’re right! I’ll double check on that! Man, so much AI news, sometimes I forget about an announcement like that. Good catch, I’ll check with them!
That's really nice! Generally, I know-- a few have asked, but I've never made a video about it, since I'm sure the wider audience doesn't really care-- haha. But, since you asked: Primarily a guitar player since my teens. Started off playing a lot of Thrash and Metal, which was great for building up speed and chops (most of which have gotten fairly sloppy over the years)-- The 90s grunge thing happened, and I transitioned into that, playing in local bands. Taste started to expand out to Jazz/Experimental. Weirdly, was playing in Punk Rock bands during that phase. Played with samplers a lot when those became affordable-- lots of MPC pad bashing and record collecting! Somewhere in there I went to film school and moved to Hollywood and did that whole thing. 17 years in the business-- but music was/is always my escape from that side of the creative world. In general, these days I'm probably a lot sloppier musically than I should be, given the years-- but, I don't get as much time as I'd like to woodshed. I've picked up enough theory that I can get through most situations-- and, I've gotten to the point where I can play about any song if given enough practice time (minus the super shreddy stuff). But yeah, these days I'm mostly home recording for fun or doing random jam sessions with my pals around town. Last band I was playing in was a few years ago now-- I just don't have the time to keep up with the channel and gigging! (Plus, I don't want to stay out until 2am anymore!)
@@TheoreticallyMedia Oh thank you for giving me such an elaborate account. Quite generous of you. Wow! You must be so busy. In a way, AI, has given you a 3rd chance in life. Thank you! But it's great you that slip in a little of your personal history, and artistic preferences now and then. It's a nice touch to the channel, to also hear about the person behind it all.
Is there any AI tool which can generate 30-40 minutes of guided meditations with music. I have my own scripts & topics. But, unable to find so far. Plz guide.
I am not getting the ‘manual mode’……..younsaid ‘instead of natural language’………didn’t understand the difference with what you put in for those scyfy jazz prompts….how were those words not ‘natural language’?……..i put in individual instruments in prompt section without using manual mode and it usually works……..can you do a something more on this?Thanks
Oh, I’m sorry- what I meant by “natural language” is more like: instead of using a sentence like “a dark cyberpunk inspired jazz song, with undertones of bebop jazz, featuring a trumpet solo” You can just write: “Darkwave bebop, trumpet” So you’re just using keywords in manual mode. It’s helpful for trying out really weird things like “Darkwave, Reggae, Tuba, Asian Drums” which is obviously kind of stupid, but might create something super unique. If you tried describing that in “natural language” the Udio model might get tripped up In words you are choosing, so manual mode is just a way of being a lot more direct about what you’re looking for. Think of it as a collection of tags for genres and instruments you want to hear.
I'll let them know. The last people I'd want to make mad are the Nords! You are all tall and strong. Probably from all the fish you eat! (haha!) Although, to note: your weird Salted Licorice is terrible! =)
Which one of the paid plans is needed the use a song to monetize a youtube video? In their FAQ it says a subscription plan is needed but which one???🤷🏻♂️
One idea to that is to download the stems, re-feed it back into Udio with a remix prompt like Acapella (voice you want) and then recombine it with your original audio, It’s kind of a workaround, but yeah, I know that’s a requested feature (along with locking a consistent voice) that they’ve been working on. I’m obviously not a dev, but I tend to think that consistency is a tough issue across all GenAI. It’s something that extends out to image generators as well.
I am trying to decide on either Suno or UDIO I want to be able to download drum stems mainly and drop them into superior drummer have you dabbled with that at all?
A bit-- it's been awhile since I've checked in on SD-- Are you now able to upload audio and have it transcribe as MIDI? On the Udio/Suno front: I use both to be honest. I lean in on Udio, but I do like Suno as well. I'd say give them both a try for a month and see which one you prefer!
@@TheoreticallyMedia Yes it has a feature called "Tracker" that separates into single midi parts Kick, Snare, etc. If the drum track is to distorted it causes a lot of problem trying to separate the drums. I was mainly curios on how good the stems were clarity wise. Anyways I think I will try it out and see what I can come up with, thanks for the response = )
Awesome request! Yeah, I didn’t dive too hard into the “power user” category here. That was something Udio and I talked about, and we’re hoping to do a follow up for you all. The thought was that going from “here’s how to prompt” to “here’s how you import a stem, quantize it, EQ it, and run sidechain compression would have been too much for most people to digest. There’s so many ideas and approaches that I want to explore though- so, hoping to do an “Advanced Mastery” video in the near future!
I'm making some songs with my voice in it with Udio, I'm also will add a guitar played by my y some of those. Is like having a band and a producer, except for my voice and instruments, so I have to achive mix them well. I don't even have to bother with messed up lyrics cause I'm doing that part too.
Sorry if I'm being bold (I don't know how the users of this AI platform deal with issues of authorship, cost etc.), but could you please complete the songs at 1:35 and 9:24 and publish them? Or at least make the full fraction you obtained from them available?
Great tutorial! I learned a lot, especially what I was doing wrong. Three questions: 1) Can you make a segment of your song just be a vocal without instruments? (ex: Bohemian Rhapsody intro) 2) Can udio use spoken words in a song without singing them? (Sibling Rivalry ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-mWKO_aYs4hE.htmlsi=cHl0yad0ForAyMyv) 3) Is there a way to make udio use a specific instrument or instruments regardless of the genre? (surf piano ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-TydFvGOHTsU.htmlsi=c4hE1a4zIc5XCQUw or bagpipes pop ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-pFr5lkuyVqo.htmlsi=Rz1RXJRIpIdG8a4A )
Is there a way to get the desired BPM? Everytime I prompt for it the song is always way off beat. I use this to extract the vocals as im a music producer.
GOOD DAY TO ALL, HAVE YOU EVER GOTEN A REVIEW FROM SOME ONE THAT, YOUR UDIO EXPORTS, AS IN TERMS SPECIFICALLY RELATIING TO THE QUALITY OF THE INSTRUMENTS TO BE SUB PAR AS IN SOUNDING OF LOW BIT RATE? DEALING WITH A PARTICULAR ARTIST I THINK JUST WANTS HIS BAND TO COVER THE TRACK AND USE IT FOR INSPIRATION LOL... THOUGHTS??
Ugh I could do unspeakable things for consistent voices like Midjourney’s consistent characters, I wanna make an album for the rockstar character I made up with MJ 😂
1000% this is a major request. I'll pass it along to them for sure...although, I'm sure they've heard it quite a bit! Still, one more vote and it might be the one that gets it done!
@@TheoreticallyMedia or hell some kinda Eleven Labs type thing where you could create a voice separate from the music! Ah, boy can dream! Def a bitch to figure out tho when you want a voices that’s Norwegian dude who sounds like a mix of Robin Gibb and Janis Joplin. Yes I am insane. 😂
Those bumpers were my jam. Kids today…they’ll never know those vibes. Ok, off to go watch Too Many Cooks. I’ll be back in 15 minutes with a melted brain.
OMG... On one hand it is incredible, yet to me it still is quite random and not at all how music should be constructed. It is as though Udio is designed by non musicians for non musicians. Do any of these AI music generator devs consult actual song writers? It would be nice and refreshing from a musician's perspective to see the dev create a generator that is based on actual music theory, like if they sat down with a professional in the old school music business, such as world renowned composers or a legendary song writer and ask them what controls they would like to have? Time signatures, Key signatures, tempo, accelerando, accidental, andante, crescendo, defined scales, chords, accents location, etc. If it were so, non musicians may actually learn something about music rather than this really confusing random bells and whistles act that is made to make people think they are in control of their creation. Here is a scenario. Lets say the user actually has a song idea, something like a melody they have been humming to themselves, and perhaps even some lyrics as well. They may even have it somewhat figured out on an instrument already like a few chords on guitar or piano. They may have already decided on a key signature, a time signature and general tempo. ( Key of Eb, 7/8 time, at 173 bpm. The user inputs these variables to the AI, and further defining prompts specifying scales, solos and adding/removing other instruments as they go. The user could actually pick up their own instrument and play their idea along with AI like it was a virtual band or bandmember, stop the session and tweak/ make corrections or take the idea in a entirely new direction, then finally record them all together for a final master recording and actually feel like they created something with their imagination and be genuinely proud like they really put some effort and two cents into the production. Hey why not have it printed out on sheet music to share with others. I'd like to see that. Please😃👍
Well, I’ll say: we’re REALLY early into this technology (like, less than a year), so it is still brand spanking new- but you raise some great points and have a wishlist similar to mine. So, I know Udio has a roster of real (and famous)’musicians they work and consult with. But- to your point: it’s sort of a horse/cart situation. The tech emerged from three developer world. And while a small percentage of them might have music backgrounds (I figure some of them have to!) the majority were more computer scientists. Suddenly AI/ML explodes and is capable of generating really stellar audio, but its “parent” is from the software side. So, what we’re seeing now is a bunch of devs talking to a bunch of musicians (New Mom?) and trying to figure out A) what they want, and B) how to implement it. To keep on with then parent analogy, we’re looking at day one of a couple moving in together and they’re trying to figure out where the pots and pans go in the kitchen and who sleeps on what side of the bed. That’s a really weird analogy. Anyhow, the tech is super impressive and is really amazing, but it’s still a fairly stupid baby. So it’ll take some time for us to figure out how to make it all work, it is getting there though and some really talented people on both sides are working on it.
Side note: sheet music would be great, but I think the overwhelming request is convert to MIDI. Which,’can be transcribed to sheet music anyhow. So, yeah- that 100% is on its way.
@@TheoreticallyMedia Thanks for the Info, good to know. Ya easy for me to say, it should work this way, and that way, yet. I really do not have a clear understanding how it all works or processes the music or art. AI Art I generation starts out as millions of pixels that get sorted out like a render, from what I learned either on your channel or another, but deep down trying to understand AI I feel like a one of the monkeys in 2001 Space Odyssey. Is the AI learning with each effort? Is it really AI? I question still. I'm an artist not a scientist, though i have a pretty good imagination and I'm an old school music generator myself. Love the videos
Good Info but still like Suno Better. Seems to select Music I am trying to create and also makes better lyrics.. But the Vocal Audio's are better udio.
I have to agree with @MistaRopa - Udio sung some gibberish completely unrelated to its own generated lyrics and inpainting made it worse! After 6 generations I gave up. I've corresponded with them about this and their reply was, "It's not an exact science" and basically, 'you get what you get' - no mention of giving me some credits for all my re-generations to fix their failures. Also, inpainting - what a piss-poor interface! You have to select a region on a timeline AND select the lyrics, which aren't easy to find when it's generated loads of crap that it hasn't even sung! Even then it often comes up with even worse gibberish or new vocals that depart waaaay too far from an established melodic signature... I've had some good results, but Udio is very hit-and-miss and has really annoyed me at times - not the right mind-set for creating, even with AI!
Thanks a lot for this very informative video, Tim. I have actually produced an entire music album titled "Interstellar Enlightenment" promoting well-being and self-care using Udio. The album features 45 tracks, Including english, instrumental and many others. I have used ChatGPT to tailor the lyrics accordingly and to do the translation where necessary. This project/ music album was also developed to show to my undergraduate students how generative AI tools could be used in positive ways in their assessments. If anyone wants to have a listen, feel free to access the Playlist: ru-vid.com/group/PLv57kSTW0t2WHpglIlTWMyvxw6KZ3UuZc&si=jVZ3J8C44ovav5-_
You may not see this comment as I am 3 days late but I hope you do. I am having issues with creating an ambient Deus Ex style music. But I have an idea that you mentioned. I will try to upload some of the original tracks and then AI gen mix them for a new type of version. That I have not done. But I noticed it just created really bad unusable music for DeusEx style. I have not given up but created about 50 songs and all of them are trash I am afraid to say.
let me tell you in the world none of any AI music generator around here who can do 2 things 1- they cant generate songs in kids voices even cant control only male or female vocal and if I want a school small kids poem they can't do always a elder adults vocals :'(((( 2- none of any AI music genator can do only vocal song without any instruments even 1% always comes with instruments so t=will u take itas challenge and gonna hands on it???
SUNO is a much much better composer than Udio: it makes POOR recordings of GOOD music. Suno shines in structure, inventiveness, coherence - but it sometimes lacks in detail & technical recording quality (worbly vocals & fake sax/piano), and is not that great for editing existing sound. Suno is kind of like Midjourney in that regard. UDIO is a better musician/audio engineer than Suno: it makes GOOD recordings of POOR music. The music Udio creates is often boring and very uninventive. But it has a few useful technical features - basically the "upscaling" of existing audio (Remix = generative cleanup), and also inpainting and extending towards the front (intros). So Udio is a bit like Stable Diffusion, which is great for image-to-image structure/style transfer or upscaling, but it's poor at creating new things. I start in Suno & then refine in Udio. And I start in Midjourney & then refine in Stable Diffusion.
Kitbashing. You are speaking my bread and butter here. Yeah, I don't think there is a "platform to rule them all"-- and in all honesty: I don't want that anyhow. That's when everything becomes homogenized. I'm actually working on a video right now about a new "Midjourney Killer" I'm so exhausted by that phrase.
What affects this issue alot are your lyrics. Suno has this ability to make catchy songs with very little vocal flaws with pretty much any kind of lyrics you put into it but at a sacrifice of audio quality/autotune sound. With udio you have to try to match the structure the particular genres you are trying to make as well as think about how much lyrics you want to generate with each model. You put too much lyrics or have wrong kind of structure then genrations will be affected negatively. I also find after update that you really have to use manual prompts as well as the automatic one just changes too wildly.
Good video. Just Subscribed!! I ve been using UDIO for 2 months ( ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-glUVJPeObeY.html) But I have a doubt: The prompt structure Should I start with the genre, the theme, the tuning. I usually start with the genre, but I don't know if that's the best course
