Today was the first day ever using Udio after being an amature producer for almost 20 years. I was able to import a royalty free instrumental with my own vocals and lyrics and then remix it and extend it to make an entirely original song using my lyrics and vocal style. It took me a few hours but being able to take those stems and put them into my DAW to then tweak was an incredible situation. So many of the sounds and instruments sounded better than anything I could record myself and was exactly in the same vein as the idea I had in the demo. I also used the vocals as a reference to then record my own vocals with the same melody. Today I created probably the most advanced and professional/Catchy song I have ever made.. 90% of it was just AI but knowing it was all built off my lyrics feels kinda good.
I love this! This is so cool to hear (read). Yeah, the remixing of audio is huge and I've been able to do some amazing things with it myself. I used it to make the music for a trailer that I'm working on that's coming out soon. You might say that it's 90% AI, but it's still you. You're still creating the lyrics, you're still putting together your ideas in the way you want to. It's a tool that's helping you create what you've wanted to hear.
Unbelievable, it’s as though you’ve been watching my use Udio, that’s my exact workflow i do, from trial & error…. Good stuff maan 💯🏴☠️🤙 We are living in glorious times, co-creation with Ai makes things possible… it’s bliss spending hours creating/designing what you are hearing, trying to improve on it… a feature I’d like Udio to do is able to isolate a particular instrument and manipulate it. Say i like the harmonica sound that a 33 sec generation produces, now I want to be able to manipulate the direction/sound style that harmonica has as each 33 sec clip … now it’s hit or miss, i can kind of manipulate it with [harmonica break] or solo, breakdown etc. i haven’t really played around with being detailed & specific in the main prompt though, which I’m going to do specifically in a few minutes from now :) Excellent tutorial for newbies though, I’ll recommend your clip to ppl when i come across who are interested in music generation 💯🤙
Darn, I hoped you wouldn't find the camera that was watching you work in Udio. ;) But yes, Udio has been great and I've been able to use it for quite a few projects and I'm constantly finding new ways to manipulate and use it. The instrument thing was something talked about early on that a lot of people want to see. I'm going to assume that it's something that they are working on by having separate tracks for voices and music. Maybe just the 2 tracks at first, but who knows? That would be very nice though. I hope your experiment yielded some good results. Thanks for watching and any sharing of the video is very much appreciated.
I’ve made over 50 creations using my own lyrics. It is an amazing tool if you personalize it. But it does take an ear and patience. When your first good creation is made you will be hooked.
I'm using udio to redo all my songs demo's I wrote over the years. some of my songs took over 150 tokens to get to sound the way I wanted out of my songs. now my songs aren't just me with my guitar and a beat machine and me singing out of key. I use vocal remover to remove the vocal out of music then I sing alone with the vocal then place my vocal on the music and have my original song with me singing properly. Now I have proper demo's witch is the way Udio should me used and thought of as a tool.
That's really cool. What are you using for the vocals remover? And it actually works well enough? I'd definitely like to take a look at that. And do you have any of your tracks online to listen to? I have heard that one of the plans for audio is to have multi-track outputs, but who knows how far off that could be. But the ability to remove or edit individual tracks would be a "game changer".
I’ve just completely recorded an old film I made 20 years ago. I could never get the score right. The composer did well but the midi sound was unusable and the orchestra attempt was very rough and amateur. I used 300 tokens and was able to rescore the whole thing and it finally looks like what I originally envisioned.
@@timefilm that's awesome! Yep, another great use of AI that gives people the opportunity to do or complete something that they otherwise would never have had the chance or ability to do. Is this going to be publicly available?
@@shadyendeavor Thank you! It will be on my channel next week, I will return and let you know. However if you go to my channel now you can see one I''ve already completed - the batman 70's cult film. I used not only udio but AI image generation also.
@@timefilm wow, that's how slow I am. I didn't even think to check your channel...ugh. But I took a quick glance and it looks like you have some great content. I'll take a look at it in a few. And subbed so I don't miss the new one.
The problem with random lyric prompting is it can lead to the meter of the lyrics influencing the music so If you are defining lyrics for a specific production, keep that in mind
Great video. I have been playing with udio for about a week now and have a few songs under my belt. The music and vocals it came up with blew my mind. My very first song is my favorite so far: kRUHLc9a84eyk4xZVmA3RF. My second song blew me away also: hsJ3sCpimLF6oytd1iwaCw. The app still needs work, though. Sometimes it skipped entire verses and added in lines that I didn't write. Still, it does impressive work.
First, thanks so much for sharing your works. I must say, I really liked the first one "Lament of the Undead." I could see that being the theme song for the undead in WoW. The second song was a bit too slow for me. I think that it might have helped with that idea that I've seen a few mention here about starting with the chorus until you find that great hook and then extending before and after it. But then again, maybe it's exactly what you wanted. All personal tastes. I could be wrong, but I'm pretty sure you said something else that you edited out of your comment that I wanted to respond to. If it wasn't you, I apologize, but to me it was something very important. The idea that you were able to create something that you knew you would probably never be able to hear without such a tool. I'm constantly being bombarded with people saying how AI is crap or it's evil or whatever. But I love the fact that it's enabling people that may not have all of the skills (or money) the ability to create what they are able to imagine. Something they'd never have been able to do before. Sure, it's disturbing job markets, but for most of us, it's allowing us to create things we never would have been able to see or hear otherwise. A net positive, imo.
@@shadyendeavor thanks for the positive feedback. I originally mentioned the fact that the song "Rise of Vecna" was based on D&D and that I never thought I would have a chance the make a song about it. Udio has given me the chance to do just that.
Thanks for this tutorial by the way. I've noticed that while making my album of robot gangsta rap, the AI started remembering not only my characters and details of their backstories, but also started remembering their individual voices, so if I have Robogangsta rapping, the AI uses a consistent voice now. Another cool thing, albeit very strange, is that the AI actually seemed to add things I was THINKING ABOUT in certain instances. This happened several times, to the point I noticed it and was a little bit creeped out. I know synthetic telepathy tech uses radio waves, so I got to staring at my modem/gateway for a while pondering skynet and the matrix.
Yep, AI is doing some crazy stuff. I'm not entirely sure it's reading our minds yet (hopefully), but there definitely seems to be an amount of anticipation of the user. If you haven't seen it, also check out The Thirteenth Floor. Good movie with similar subject matter that was completely ignored because it came out around the same time as The Matrix.
@@shadyendeavor Ahh yes. I've seen the 13th floor. Have you heard of synthetic telepathy? It's an old technology that uses radio waves to project a voice directly into a person's brain. Imagine what AI could do with it, using our modem/router/gateway. Brain-Machine Interfaces are becoming a big thing in recent years. I just wonder how far it's gone without the public knowing.
I'll be honest, it's one of those things that I keep meaning to try out - putting in lyrics from an already well known existing song - but I always get too busy with other stuff. But yeah, I can imagine a song like Rock Lobster would have some odd outputs. I'm assuming the "no prompt" thing is what made it go crazy and pick out some of the lyrics to guide the music.
The technology is a good start. The music indutry is not going to be shaking in their boots. I can imagine better pro versions of the technology sometime down the pike with far more capabilities. As it stands right now, it's more of hobby pastime kinda thing to playwith than anything truly useful. I can say one thing for certainty, that if you do not like music as it is today, then brace yourself, you will be getting a lot more of it. Lots and lots more of it, This type of thing is going to be churning out tons and tons of that slop non stop. If you see yourself as a music conesuir, consider this an introduction to your vision of hell.
So if you believe that this is just going to be "churning out tons and tons of that slop non stop", how is that not going to adversely affect the music industry that was doing that by itself until now? Sounds like slop competition to me.
@@shadyendeavor Good point.The music industry on the corporate side has had a lot on their plate manuevering the ever changing technological landscape even before AI, but they are experts at marketing slop, they'll be fine. On the creative side, artsits have also not been happy with the current model for quite some time even before AI. Nothing changes that. Also on the creative front, I think it may incentivize new artists to up their game, especially as live performaces have been the go-to model to make a living on music. AI is just a tool, and as alwasys, young artists are going to find a way to use it creatively. I think it will help them realize their ideas better and faster and hopefully create new takes on music. Ludites for a brief moment go after the hew music loom, egged.on by faux outrage on social media such as we saw for the new iPad Pro commercial, but life goes on. Amazing technology.
Thanks for this great and helpful video. I'm new to Udio, but very family with audio and video programs, so I'm usually pretty quick to figure out new things like this. However, I'm having trouble getting male vocals. I've created custom lyrics, and using Country/Rock in the top filed for style, I'm assuming that's the prompt box. I have [male vocals] in the lyric box, I tried it first at the top of the lyrics, then dropped the tag above and below the [Verse] tag (separately) but I'm still getting female vocals. Is it because the original version of this certation put out female vocals and you can change that after the first version is created? Currently I'm in the "Remix" mode trying to attempt this tweak. Any help would be appreciated. Thanks
Update: I was able to get male vocal after sliding the "Variance" slider all the way over to Different. This changed the melody and music completely, but I got male vocals now. Maybe if I didn't slide it so far it would have kept the feel of the song. Might try that. I like the music before, just wanted male vocals. Thanks again.
If you've already created the first part of the music and it has a voice in it and you're remixing off of that, then yes, it's going to be difficult to change the voice at that point. But if you're just talking about having made something and you're not remixing off of it, then you should be able to change it. I definitely went through many many iterations until I was able to get not only the voice that I liked, but also the music to go along with it. As many people have suggested (and I will be adding to my next video on Udio), a good idea is to start with the chorus (or maybe even the bridge), so that you already have a good feeling for the entirety of the song, including the voice, music, tempo, etc. But still, just putting in "male vocals" should give you, well, male vocals the majority of the time. So I'm really not sure why it wasn't. But AI tends to not give you what you ask for a lot of the time. In fact, with a video ai that I was working with for another video, I literally had to type "waking AWAY from the camera" in order to get the character to walk toward the camera. I'd say it's the "ghost in the machine", but that might be a bit too on the nose. If you like the music and you have a part that doesn't have singing in it, you could try a new track and start by uploading that part and build off of it instead. Just an idea. That might give it a clean slate for the vocals.
I want to make a song with more than one vocalist, like a duo/collab/feat. Want to one singer sing most of the song and the second one to do specific verse/chorus.
When I get around to making another Udio video, I'll be sure to discuss everything that I can think of that is not in this video. I don't believe either of those features were even available when I made this video. Though I could be wrong.
Sometimes I will add a chorus first and it will be the intro to the verse, other times if I think the verse is strong I will let it be the song guide. I almost always use the intro. since I write most of my lyrics these methods work for me. but the important thing is you have to experiment...and like anything else...practice...practice...practice🎶🎶🎧🎧
Yep, I've seen this mentioned before. I haven't tried it yet, but I think it's a great idea. Even to try it with the bridge instead of the chorus. I even mentioned it in my newer, much shorter version of this video that came out a day or 2 ago.
I tried Suno and Udio, but I had a song I had made in FL studio 9 long ago and lost in a hard drive crash. Udio made me a better version of it, and I was blown away. I've dropped $50 on Udio already, buying credits. I made a 30 track album of robot gangsta rap, and now I'm 10 tracks deep into an album that's metal and electronica, called Tales of Woe & Wonder. Each song tells a story. I freakin' love Udio. I haven't done much with Suno, though I wasn't unimpressed with it. I just stuck with Udio because it remade a song I thought I had lost forever. That song is on my channel if anyone is curious. It's called Strange Little Bird.
Yeah, Udio is amazing. I'm working on a project that I've put well over a hundred hours of work into that should be up sometime this week and I scored the audio using Udio and I couldn't be more happy with it. Suno definitely has it's use cases too. From what I've heard, the prompt adherence can be better with Suno, but I think just the quality of the output with Udio is better, but you just might have to work at it more. And now with 1.5, it's even better.
@@shadyendeavor I've tried Suno, and I agree Udio is better imho. I finally got my 30 track album "Autonomous Origins" uploaded, and now I'm 10 tracks deep into making a second album "Tales of Woe & Wonder". I have 5 tracks from it uploaded so far. Seriously, with some of this music you cannot even tell it's AI.
I have been at it a week now and love Udio so thanks for this video. I have also been pouring through the subreddit and will check other forums to help with these issues (I am using Udio to output instrumental music only no lyrics) (1) How can I prompt udio to REMOVE things like guitar solos or unrelated genres? “no guitar solo” is often ignored. I have also had some success using custom lyrics then not entering any lyrics or just (.) in first line. FEATURE REQUESTS I SENT TO UDIO TEAM: (1) A custom template would be a good idea. Then I wouldn’t need to delete random genres that are generated it would stick to my template. Even when I use the same prompts this happens consistently. (2) Restrict output to key signatures or related modes (I can enter “F Minor” but it’s hit or miss) (3) Allow audio input to detect key signature and generate extensions from custom audio (I have tons of music snippets lying around that are just unfinished ideas) (3a) allow chords as prompts (similar to above but more specific “a song using C Major E minor and B Diminished”)
sadly, as far as I know, there is no "negative prompting" yet for udio. Only thing I could think of is if there is a term for types of music being played that automatically have a specific instrument excluded. Just like by calling it "a capella", that should automatically exclude all instruments. So is there something similar that excludes specific ones? Dunno. Just a thought. Or maybe something weird like "silent guitar"? The requests, though they sound great, are all probably things that will eventually make it into it, but are currently too complicated or just lower on their priority list. I know quite a few people have asked for audio upload references.
Currently, the AI doesn't seem to understand the rhythmic interaction of drums and bass. The AI is also currently unable to give an instrument an independent counter-voice in parallel to the voice. Apparently the AI does not know that the overall sound consists of individual instruments that have to coordinate their individual melodies with each other. But 99,9% of what the music industry produces is hardly better. Quantity instead of quality. Therefore, AI will hopefully destroy the boring mass music and promote the good artists who offer quality.
Heh, exactly. Yeah, it's far from perfect for people that really LISTEN to music. But the mainstream pop music industry is in trouble. Real, good quality music is still in the hands and minds of the artists and audio can just be another tool for them.
Great video!! I wish this was around when I was getting started with Udio/Suno. I was skeptical of the word “Masterclass,” but it is well-earned! The final version is fantastic. Here’s some additional tips I'd add: 1) re: Using 6 lines / 250-300 characters per generation, if you use more, it will try it's best to squeeze them all in (which is great if you want fast rap music), or use less if you want things drawn out, like in a soul song. 2) For prompting, try putting ‘jhdfg’ (any random text). The genres it will spit out in the generated clip are pretty off the wall. A lot of ‘experimental’. 3) Sometimes I’ll notate the title of the song with “goes off the rails at 2:21” or something like that, to help me keep track of the clips I like but don’t love. 4) Hitting cmd - a couple times to reduce the font size in your browser can buy you a lot of real estate. Helpful when you’re dealing with long lists of tracks. 5) For instrumental Outros that actually END, I had great success with |-> Add Outro, ( • ) Custom: and pasting this into the lyrics: [End] [Exit] [Fade Out] [Finish] [Outro] [Quit] [Silence] [Stop] 6) Make sure you download a WAV file, rather than MP3. WAVs use lossless compression and are higher quality. 7) At 36:47, you have a great opportunity to demonstrate Inpainting. There’s a loud, abrupt “ooh” that doesn’t belong in the song.
Yes, I was skeptical of it myself. ;) I just knew that I wanted to make something that completely showed the entire workflow of an entire song and I think I got it to that point. Yes, there were definitely quite a few things that I missed. But this was very early on and Udio was still in beta, so I, along with everyone was still learning at the time. Definitely some good tips and if and when I create another video on Udio (especially with 1.5 available now), I'll probably add most of these along with quite a few other things I've learned since. And yes, the inpainting would have been nice...had it been available to me when I made the video. I think it was released a day or 2 before I released this video and I wasn't going to go back and "fix" it when I hadn't even paid for Udio yet. Anyway, thanks for the comments and suggestions. Like I said, I'll definitely look at them and work on them with the next video.
I like to enter my chorus first so I get a better idea if it’s going to generate a strong hook. Then I’ll add the 1st verse before it, and 2nd verse after.
Good beginners guide. I was all set to make some beginner tutorials now that new features have dropped but I see you got this. Inpainting in particular is causing a lot of confusion for first timers. I will past a link for this video in the UDIO discord to guide newcomers. Look forward to seeing more like this.
Yep, as I was just about done editing this right when they announced the paid plans and in-painting, I, obviously, didn't have time to add information on that into the video. But I'll probably work on an addition to this soon that goes over the in-painting feature.
@@shadyendeavor ... OMG... This is really strange, dig this. I had spammed your video literally 150+ times with this comment, as they were being removed within seconds. The LAST time I did the same thing on another video, the one comment that finally survived was "double-pasted", just like this one. I had another take that was removed, but will try it again, to see if it works now.
@@shadyendeavor ...The song you made will satisfy the vast majority of pop-radio and Billboard requirements and for a MASSIVE number of people. Been an absolutely MONSTER music fan, since like 12 yo and this is as good as most of what we see on top 100 pop-charts. I would agree... the music industry is in deep trouble with this technology. Serious music fans will still demand human songs, but this song you made will fool far more people than serious music fans would ever be comfortable in admitting.
@@NinjaKittyBonks Just to be clear, are you talking about the "original" version of the song or the one more "country folk" version at the end? or both? I think the "country folk" version is better and therefor would not do as well on the pop-charts. But yes, considering what pop music has become, it doesn't take much for the populous to, well, make it popular. I have found that some of the pop music that is out there, though crap, can actually be pretty good when it's covered by someone with competence and actual musical skills. There have been quite a few times where I've come across a song I really like and didn't realize it was a cover of something terrible like Gaga or Bieber, but the cover musician was able to take the good parts of it and making something much better out of it. Kind of a tangent, but whatever.
@@shadyendeavor From a logical point of view, the more syllables used per line, the higher the risk of the ai messing them up (wrong flow of words, emphasis on wrong syllables) I've tried a couple of times to ask the chat ai to align the number of syllables per line and it did work to some degree, but not 100%. Generally, though, the lines need a good rhythmic flow, as fool-proof as possible, and it may be a good idea to try reading out the lines before entering them, just to make sure the flow is good and natural.
I asked claude on 6/6/24 to write me a song and received this. "I'm afraid I cannot reproduce or create full song lyrics, as that could violate copyright laws. However, I can provide a general outline or summarize the requested elements without including full lyrics"
I literally just tried it five minutes after your post and it worked fine. Did you describe specifically what you want the song to be about? Or did you just give a vague idea of the type of song? Did you include any names of artists/groups?
Excellent tutorial!!!!! After watching all of the iterations that you had to go through, that’s why they are generous with the number of monthly tries they allow you to use. My only concern with all of this is that since the lyrics are mostly computer generated, I think songs will lose that human experience touch.
That depends on how you look at it. First off, I created my own lyrics, though I used Claude to get some ideas. And then, because of that, that is why I believe quality music will continue on as people with a good ear can tell the difference and seek out the high quality music with good writing. However, the mainstream pop music industry is in trouble. That's how I see things going - at least for the near future.
Currently, most all AI, you are not able to copyright. However, you are able to use the songs you create for commercial use according to Udio and their terms of service. So there is definitely no problem using them on RU-vid. I don't know how it works if someone else were to use your music though.
@@shadyendeavor yeah great, im not going too.. hey! I was thinking about your next video idea you could do.. there's a new AI video creation called clipwise ai. It basically make your entire short youtube video in 3 minutes.from just your prompt even adds text to Flow with words..ai generated voice, . free download for 2 videos. Watch a clipwise youtube video. I was impressed
I've used Udio quite a lot now, and in my experience, by far the most frustrating problem is when the program downright ignores even high priority prompts or - even worse - goes directly against them. Some examples: was working on a love song to a girl, so used "male vocalist" almost at #1 position of tags. But Udio insisted on giving me extremely female vocals again and again and again. I think I made 30 generations, and out of them, only 2-3 had male vocals (thankfully one of them was perfect, so I did get the song made) Another example: trying to end a song, so I chose (outro) and entered the final lyrics... Udio completely ignored it and generated a standard clip leading up to another section. I tried this over and over, with the same result. Then tried using the prompt [chorus to fade] and/or [chorus fade], which was also completely ignored. In the end I gave up and took a break, assuming that Udio was temporarily braindead. When I returned later and tried again, Udio produced an absoulutely astonishingly beautiful ending in the very first attempt! Another common problem is the use of bridges. Almost every time, Udio insists that a bridge is just another word for verse. Using "advanced controls" and adjusting the slider simply results in Udio creating a bridge that has no connection whatsoever to the rest of the song and not only that, it also forgets the verses and chorus from that point, as if you're starting a completely new song. The reason these problems become so frustrating is, that no matter how horribly your prompts get ignored, the generations still count and it's VERY easy to burn 30-40 credits just fighting against the algorithm before you get something you can work with. This is not just time consuming, it can also get very expensive and it feels so unnecessary.
Quite a bit to respond to here, but yes, for the most part, I agree. First, as far as the vocals go, yes, I have had this problem a lot. One thing that worked much better for me is to not rely on the prompt and instead put it into the lyrics. Do not put it in the same brackets as verse, chorus, or bridge, but on a separate line and you can even do multiple different ones during the same creation [Verse 1] [Male Vocalist 1] We were just two kids, from the poor side of town We had nothing but what we haaaad in each other [Male Vocalist 2] Together we dreamed of so much more more than this struggle more than this strife [Female Vocals] I promised myself, one day, I'd give her the perfect life Just as an example. It doesn't always work, but it seems to contribute quite a bit. The problem you were having with the Outro WAS a known bug. People like you and myself were reporting the issue and it was with both regular extensions becoming Outros and Outros just becoming regular extensions. This was posted on their Discord on the 20th: "We have fixed a number of small bugs related to extensions including issues with unintentionally creating outros and cropping the first fraction of a track when using trim mode" I haven't seemed to have the bridge issue, so I can't really say much on that. Either way, yes, there are definitely some very frustrating results that we get, but I still think this tool is amazing and it's only going to get better. I'm sure that at some point we'll be able to literally ask specifically for, not only a male or female voice, but a the specific timber or gruffness or whatever too. Just as I think we'll be able to get more specific with the instruments that currently don't exactly follow our requests perfectly either.
Thanks for the tips, some I already knew, others I learned, it's a little difficult to get something good now, I subscribed to Udio, before when it was free I got better results. The three songs were made with the free plan. Warriors of the World Prompt: rock, heavy metal, power metal, male vocalist eGeDfr6Bc49A6i4uzJ494e Reminiscent of Love Prompt: R&b, Contemporary r&b, Neo-soul, Soul, Pop, Pop soul foGKuJW7Ne6nxtTPavgutA Canto Primeiro - Poema 4 - v1 Prompt: classical, romantic classical, violin, fado qdUrt6NnDwbkiFdT7jhpaX
Very cool! I'll check those out a bit later when I can turn up the speakers. ;) Also, without having heard them, I love that all 3 (based on their names and prompts) are completely different styles/types of music. I honestly can't say I've noticed a drop in quality, but to be fair, I also don't use Udio nearly as much as I did while it was free. Kind of ironic that now as I'm paying for it, I use it less. But what I have created since, I've still been very happy with.
Is this a problem that is still happening? Or was this something you last tried a couple of days ago? As this was a known issue and Udio supposedly fixed it a day ago or 2. If not, I'm not sure what the issue could be. As long as you have the Outro selected and in addition, if you're doing custom lyrics and put in [Outro], I have no idea why it wouldn't work. At least a couple times it should work.
Yep, completely understood. Every once in a while something will sneak onto my radar that is decent, but not often. You can definitely create non-pop stuff pretty well with this. I made one track that sounded like Steve Vai playing for 3 minutes straight. Sounded pretty good. So yes, maybe AI is what will actually "save" music in some aspect. It still takes human ingenuity of some sort, so I don't feel too bad.
now we can choose to truncate or only keep a section in the middle to append, big qol improvement. also, spamming [bridge] is an easy way to get good variations : ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-XHcIpb5Q61E.html
yeah, the feature, like the paid plans, were announced during my editing of this, so I wasn't able to add that in here. The [bridge] idea sounds like the same idea as what someone else mentioned here, but instead using [chorus]. I can see how both ways would be good ideas to try out. Probably depends on the lyrics you have for which would be better.
