I had to come and leave w comment. I love your reaction, and to be fair.. the music was REALLY DAMN GOOD. I love SUNO & UDIO. They have their own unique styles but to be blunt.... UDIO is far ahead of SUNO, the songs from UDIO have a unique quality, different from SUNO. lastly, UDIO is more clear (or was) and I haven't played with 1.5 yet
The Stem separation is after The song is generated. It's the same thing we can do with other software. They need to train the AI on separating the instruments as it's making the song like being tracked in a recording studio.
True. But in my brief experimentation Udio seems to do a better job "stemming" Udio songs, than some other apps are at stemming Udio songs. He points this out around 12:35. I hope that makes sense.
Man… The look of genuine disbelief at how good that first Candy iteration came out.. says it all. This is next level stuff. Thanks for all your hard work, mate.
That was exactly what my reaction was to some of the songs I created. UDIO is now better than most artists. It will soon overtake the top talent in the world. Just a matter of time.
Talking about Suno (what I use) I think the app wasn't made thinking in musicians. Cause some important settings don't exist and if you use them in the prompt they are ignored. Example: If I say I wanna a 115 BPM track... Ore this track in Bm... It's a smart decision in the start cause if many people can use it easy, the "merchandise" work better. But I also think that more and more this options will be implemented. Tks for always show us the more fresh news about this world that will change the world. Update: After testing Suno Stem Separation I realized is not good yet. My Teory: Udio implanted it in a planned way. But Suno just didn't want to be behind and released like they could. So ins't so good like we spected.
Suno usually follows the BPM for me. But it depends on the genre more. If you use genres that are usually fast, then it will usually ignore the BPM and still be fast. I've noticed the genre seems to dictate the most, even for instruments used.
@@Liminal_Life I understand if I ask something unusual in some style it will probably ignore. In fact it is the exact opposite in my case. I create the same style every time and sometimes the song is too fast for that vibe or style. I try to use the BMP to force it to stay in the standard. But just work sometimes.
I"ve been watching your videos and enjoying your content for a few months now. This video was unreal! I'm making my morning coffee playing it on the TV and my wife comes into the kitchen, she can't see the TV in the living room, and asks... "What play is this from?", then asks why I look so perplexed. I then walk over into the living room and see your face, and we're making the same face. What a time to be alive!
Thanks Bob I just subscribed! You are a very admirable person. You don't need to worry about being long winded or playing full songs I think even if your video were 4 hours long I would still watch!
I watch your vids and other AI review vids, and I gotta say: You have the best results of anyone else by far. I think your prompts fit the AI better than others. It's your music background, I bet!
Really enjoyed listening (and watching your reactions, Bob 🙂). Delighted you're enjoying v1.5, and kudos to you for your creative crafting and thoughtful commentary!
@@UdioAdam Hi! Can you give a quick summary of where we are with the law suit junk. I love Udio and generated some good songs but am afraid to do anything with them. I mean are people copyrighting these songs especially when using their own lyrics? Are people putting these songs on CZcams, are they distributing them through Distrokid, CD Baby? Where the hell are we? Ha!
Stems is not the most important update for 1.5. It is Remix. You can change a song completely with the new remix feature. I wish I had your video and speaking ability to do a video but I don't. So I point this out so you can check into the updated remix feature and may consider doing a video just on remix. Believe me you could.
Glad to see these things improving already! Honestly, my only hope was to get the vocal stem from the most recent song I’ve worked on. A 12 year old song that had my own instrumentation but no vocals. I used Suno to get a sweet vocal track for it and I extracted the stem in FL Studio and I’m extremely happy with everything about it other than the quality. Sounds like a chorus sort of effect. So I was hoping I’d download the stem directly from Suno and maybe it would be a massive improvement. But I’m dissatisfied to find that it sounds exactly the same
1.5 is really good but I find it generates a TON of hallucinations. In my experience you have to play with the lyrics a bit or use the trimmer thingy to get the timings right. If the timing of the lyrics is off, it throws a wobbler.
That would be a nice project... Remix in a DAW using EzDrummer and Ezbass and maybe Scaler to figure out the chords. These stems sound like they're actually useful. Overall sound quality has really improved a lot.
I wish you could update lyrics only on remix. So often I want to tweak a line (like class vs glass) or spell a word phonetically to get a better sound.
From my experiments on the Udio Stems : 1) are more fast and confortable to get than external services 2) are way better quality than any external service I tryed. All of them had the bad habit to downgrade the bitrate of resulting .wav (my Cool Edit regulary warned me about) Also, with Udio Stem I didnt hear that "hoovering" effect on human voices. So Udio Stems are really a good step forward... BUT in some cases, I also heard a clear voice echo in the "Other" track. This means they are not perfect and NO, unfortunately, in most cases you cant have the "Instrumental" version of your song just mixing "other+bass+drums", because some voices will be still around...
With all of the AI swirling around I feel like I have joined the land of the luddites with my #HumanMade music. Sigh... But I will admit that Udio has become most impressive. As soon as the ethical issues are sorted and more stem options are available Udio will be an amazing tool. Suno is playing catchup it seems.
Human made will always be special. I mean, we don't use GPT to do our journaling for us, do we? Human made music will always be the most personal and therapeutic. I love AI music, but to me it's like an abstract painting where you throw paints at the canvas and rejoice at the serendipity of what you get: it's fun and amusing, but it will never be as personal and therapeutic as making "natural intelligence" music.
Look at it this way, there are a lot of people out in the world who have always wanted to make music such as myself, but I was never able to learn any instrument very well for whatever reason I don't know. Classic white guy syndrome, no rhythm. Now I am able to go back and get all of the music I wrote years and years ago and bring them to life, with the help of AI.
@@TheAceTroubleshooter That's awesome, and in itself it's more than reason enough for AI music to exist and thrive. I'm just speaking from the perspective of a musician (one who doesn't despise AI, much the opposite of that).
@@spadogs You know, our point of view is the minority, I think. I'm seeing a lot of vitriol towards AI from musicians and artists. I don't think it's warranted, though - like I said, human made art will always be important and special. I agree, the IP issues need to be sorted out. There are not infinite melodies though, it's a combination of a finite number of notes. As the amount of songs being created each day skyrockets, it's inevitable that many will sound like existing ones... not sure how that's going to be addressed.
Something I love with stems is how you can split out the vocals and then put them through another AI to change the voice. But what would also be great is if you could take an instrument say the guitar part and use AI to turn it into another instrument like piano, or saxophone. But have the melody and timing etc stay exactly the same. That would be incredible and really open up remixing songs.
Well, to some extent, you can do things like that with Kits and Audimee, both of which can convert various sounds into instruments, as well as RipX which gives you some ability to replace the instruments in the stems. None of it is perfect though.
I have found it hard of late getting modern sounding tracks out of Udio. Had put it down to them maybe removing a lot of content from its training data.
Udio has never been great for modern music but it did excel at older music until they changed the model, now its ridiculously difficult to get anything usable out of it on 1.5 or the original 1.0 model. Suno is far better at modern music but sounds too robotoic for me and the audio quality sounds like a bad MP3 rip.
@@maccagrabme That's pretty much the long and short of it. Udio has MUCH better audio quality and natural sounding voices, but is also much less reliable in making good musical choices. 9 out of 10 are completely off the wall and nearly unlistenable. Suno, on the other hand, can crank out very listenable and very enjoyable music almost every time, but the singing voices are very noticeably "robotic", except for the rare exception. If we could just get Udio voices inside Suno I'd probably be very happy.
@@TheWayOfJamesBandI get very good voices in suno. It depends on the genre and extra prompts. The common example is "soulful" RnB, instead of just RnB. I also get good results from rock and emo genres. But it depends. 1 out of maybe 10 generations is quite good. Others can be average to terrible.
That is mind-blowing.And thank you so much for your efforts in exploring and explaining this AJ stuff that's really not so easy to grasp for guys like me (65+)!
I want someone to come out with an A.I. app that lets you load a song and it accurately spits out the music notation like the chords for the song the melody and more in MIDI form or something similar.
Getting optimal results with Ai stem separation usually involves trying out a few model variations and picking the best one. They may need to diversify by genre to get the best results.
Having used it for 2 days, I’ve found 1.5 sounds better, yes. But I’ve found it a bit harder to prompt fused styles, and off-genre electronic, for some reason. Especially in the 2 minutes outputs.
About the video... One of my concerns was... It will be a stem separation or a stem download option. I mean. When the aí make the song it make it "mixed" and after remove the vocal (like we do) or it make de song separately and we can download one of this? My opinion Udio have the stems before the final "mix", or the separation system they use are so much better than Suno(based on your examples cause I didn't test any of them). It seems like the Suno voice stem have the same aspects of a separation made on Vocal Remover, Moises or another service like this. I'll test, but what you show gave me an idea what I can expect. 😊
Love it, Bob. You could show us how to separate stems in Udio and change the instrumentals like the bass and drums and mix it with your own bass and drums :)
Do you publish any of your Udio gens? if not, you should start with that first candyland song; I'm sure that'll trend, it was great. You can't get duets like that consistently with one prompt.
I wonder why no one has created a stems outpaining for stems quality meaning that it would take the dirty digital sounding stems with artifacts and predict how these should should as clean signal devoid of those glitches. It should be super easy for someone who knows what they are doing to have a dataset if true stems vs ai generated stems for the same songs and then create a engine that would provide stem cleaning.
I recently psuedo-"upscaled" some noisy suno songs using the Remix function on udio with a low variance setting 0.1 to 0.15. It got rid of almost all the noise and rebuilt the vocals and instrumentation, keeping it very close to the original with some minor adlibs, absolutely nuts. Though this was done on a whole song, I plan to test it further on a noisy vocal stem with "acapella" in the prompt
@@MCG3333 that's mighty interesting. Can you explain the workflow in detail? I'm more of a Suno user but I tried applying a low pass filter to a stem I needed with hopes it will treat the provided sample as a fiktered-out intro to the full melody but it just provided songs that had filter applied throughout.
Hi Bob I tried this function I find it far below average it seems that there is a kind of phaser on the channels or stems that I used to use in the studio on my guitar It is still far from Studio quality
@@rubekker5572 my experience is that results vary. For the most part I get really nice, clean stems, but I think it’s ultimately gonna depend on how complicated the mix is.
Great channel. I find it very useful and informative about these 2 programs....possibly might make me switch to Udio! Love your enthusiasm! I think Udio seems to give much more realistic vocals than Suno, who often sound ever so slightly vocoder-ish.
@@andrewmartin8565 The will help you make the same version of the song. So if you liked a song made with udio 1.0, you can select it, put it in as the seed for a creation with udio 1.5, copy all the other settings (style, lyrics...) and you will get a improved version of that same song. Or you want to use other lyrics, but have udio use the same music, you also can use the seed.
I've been using Suno for a while now and are quite happy with it. However, where it shines in some genres, it struggle in others in terms of quality of the different instruments being used. The sound quality of a flute for example, can in many cases sound terrible because of the poor sound quality, same goes for the female vocals sometimes(or any high pitch sound really). I wish it was possible to choose different vocals from both female/male
You can control the voices by prompting "sopran" or "bariton". I always use "bariton" because the lower voices sound more realistic with less artefacts.
Wow fantastic! I was wondering, does the AI also create the videos that match the lyrics, or did you create it later with another tool? if the latter, what tool? Thank you, just subscribed.
I can SO relate to your expressions on the first gummy bear song! These moments are the best, jaw dropping. And big companies want to take this away and control AI music themselves...
Suno is better than Udio, both are useful for real musicians, and both suck for UI design. Once your library is large, going back to find early works, well, be sure to tell someone where you’re going first, because you never find your way back. Nice to see this update improves Udio’s vocals, which garble text often.
is there any AI software you can put the lyrics in and your own melody! because i would love to hear a choir sing my melody and lyrics. i know there’s band in a box but it won’t do vocals choir thanks
thank you for the video, i was curious on how the stem separation on suno competes with the other free Ai stem separation .. seems to still have vocal bleed plus it does not separate tracks.. im gonna hold on subscribing suno for a while more
Iblove your reactions because I can tell they were totally authentic. It is obvious you have a very sensitive soul, your non-verbal language always tells the truth. Very touching video Bob! 🫸🏻🫷🏻😎
Wow UDIO is the clear winner of this,, I already tried them stems seperation on suno before i watched this and after just 1 minute of playing around with, i knew it was unusable. I mean considering its AI, i expected high quality vocal stems, not WORSE quality than the vocal seperator. So my biggest question is "Why add a feature that is unusable" Same thing with their new design,, is it just me or was the old design much more,,, you know,,,, good? But after watching your video, the UDIO song AND stems was so high quality and i think i was just as surprised as you, i mean, just wow... So guess it´s time for me to switch to UDIO untill suno steps up their game. So thank you very much Bob for sharing this, i enjoy your videos and are always looking forward to some new exiting AI tools to use in my music creation, thank you for being here ;)
Hi! Crazy stuff, nice job. Do we have any idea where we are in legalities of using this stuff, distributing it. I'm not sure how to approach this without knowing the rules.
It's unfortunate that this video doesn't highlight why so many people are actually unhappy with version 1.5, due to a lot of instruments sounding like MIDI and unnatural vocals. You got a good generation. Not all of us do.
I was excited to see that Suno has stem separation, but they didn’t do it, right…. I can still hear the AI vocals bleeding in the instrumental. I thought they were going to make it to where the AI has his own track and the instrumental has his own track but literally separate with his own program instead of it separating the vocals in the instrumental like every other AI does it if that makes sense…. They should have it to where it automatically does it from the start meaning if we generate a song, it automatically releases two tracks, the a cappella, and the the instrumental without us having to use the AI to strip the vocals from the instrumental 😂 I hope I’m making sense
I agree, that would’ve been great. But, on the other hand, I think it would also be extremely tricky to pull this off, because they would need to have access to the multitracks or stems, which they almost certainly don’t have in the vast majority of cases.
I agree, that would’ve been great. But, on the other hand, I think it would be extremely tricky to pull this off, because they would have to train an entirely new model, and have access to the actual multitracks or stems, which they almost certainly don’t have in the vast majority of cases.
If you take a screenshot of yourself with your average lighting setep, then feed that image to an AI and let it create this type of background for you, but then hopefully the lighting will look more coherent and it will look more realistic, or even just striahgt up cool. Of course next you have to cut yourself out using maybe something like photoshop filler of empty spaces on a pictures. Or maybe another IA can do that to.
I think I asked this before. Not sure if I got reply or not. I notice the videos showing as you were listening to your Candy Land creations. My question is, how do you generate those videos? Are they part of Udio or did you use a 3rd party tool to create the videos? Thanks
I have a challege for you.. With AI (udio,suno) make an intro (or two) Like a DJ Intro or a like a enter the ice intro for hockey. Is it possible, presentation voices, epic music, lots of tags to create the perfect one :) Are you up for that challenge?? (And ofc it make me very happy if you are able to make it good! )
I prefer Suno,in my opinion make better songs and more catchy,and in less time,but the quality of sound is little poor. In my opinion Suno must improve in this.
As the most music AI that I know: A question. My wife sings really well, and has composed several songs. Isn´tthere an AI that the input is a capella song and the output is the song with all the instruments? All AI that make stems has the resouces to train one AI able to do that. If it appears, pls tell us about it. And Thank you so much for you teaching videos.
What you'r saying in 13:14 is wrong, cuz it splits the Stems after generating the whole song, like many AI or even Logic Pro 11 does. If you be a professional producer you know that this kind of separation can never be used in a professional world, because of lack of quality. UDIO engine makes the song in whole, not like a DAW in different channels or Stems. I know it's gonna do that eventually but I believe it's gonna take a very long time up to that point, cuz their users are not pro so they don't need it, and the Company already make a good profits, so no need to push. BTW these Stems are already processed, so nothing you can do with them when you add them to a project in a DAW.
Suno has overly compressed voices (which can be nice _sometinmes_) compared to Udio, but has a way more versatile output. It adds bridged on its own, quite surprising musical ideas to chorusses, that like. Impressive! However, the stem separation in Suno is not really convincing, when it comes to refine that song 'at home'. Well, my experience is all based on your videos, yet. I am going to get my feet wet soon.
think uido wins on the stems.. i found it is hit and miss with suno and also its not multi split like suno its just vocals and song.. and you can still hear vocals in the background.. but udio Mmmm sweet
They both make music that sounds ok, silly kinda nursery rhymes are not the same as radio-quality catchy pop songs or rock songs that could be a hit, can they make music that's really good that's where the frustration starts because they can but its a long road stumbling around in the dark, controlling what you get is random where right now we are monkeys pulling a lever hoping for a treat as there is no concise manual on what prompts deliver that is put out by developers, they both need manuals on what prompts do they must know how to talk to their software, imagine software like Photoshop with no manual just a line that says push buttons until you find something you like, oh and pay for every random experiment where 95% will go in the trash..
Depends on the genre and luck. For me, if audio sounds low or bad on more than one generation, I wait and reload after a while. I feel like they have people working shifts, and each person adjusts the EQ different. 🤷
For anyone who may care, I finished out "The Story of My Life" www.udio.com/songs/k9mGHNGcJxrk5XemnYjff5
I had to come and leave w comment. I love your reaction, and to be fair.. the music was REALLY DAMN GOOD. I love SUNO & UDIO. They have their own unique styles but to be blunt.... UDIO is far ahead of SUNO, the songs from UDIO have a unique quality, different from SUNO. lastly, UDIO is more clear (or was) and I haven't played with 1.5 yet
I liked the other generation better
The Stem separation is after The song is generated. It's the same thing we can do with other software. They need to train the AI on separating the instruments as it's making the song like being tracked in a recording studio.
I noticed that too, it was like it was using AI separation after the generation.
i know what they made is bullshit
True. But in my brief experimentation Udio seems to do a better job "stemming" Udio songs, than some other apps are at stemming Udio songs. He points this out around 12:35.
I hope that makes sense.
You would need actual AI for that not ‘just’ extremely good algorithms… my limited brain says….😅
Can you explain more. Are you saying we cannot separate the drums into their own track so you can throw it in a Daw or what am I missing?
Man… The look of genuine disbelief at how good that first Candy iteration came out.. says it all. This is next level stuff. Thanks for all your hard work, mate.
Thanks very much!
Yeah i subscribed to this channel based on that moment lol
That was exactly what my reaction was to some of the songs I created. UDIO is now better than most artists. It will soon overtake the top talent in the world. Just a matter of time.
Bob your disbelief is the exact reaction I had when I used Udio the first time. I’m so excited and cautious
Talking about Suno (what I use) I think the app wasn't made thinking in musicians. Cause some important settings don't exist and if you use them in the prompt they are ignored. Example: If I say I wanna a 115 BPM track... Ore this track in Bm... It's a smart decision in the start cause if many people can use it easy, the "merchandise" work better. But I also think that more and more this options will be implemented. Tks for always show us the more fresh news about this world that will change the world.
Update: After testing Suno Stem Separation I realized is not good yet. My Teory: Udio implanted it in a planned way. But Suno just didn't want to be behind and released like they could. So ins't so good like we spected.
Suno usually follows the BPM for me. But it depends on the genre more. If you use genres that are usually fast, then it will usually ignore the BPM and still be fast. I've noticed the genre seems to dictate the most, even for instruments used.
@@Liminal_Life I understand if I ask something unusual in some style it will probably ignore. In fact it is the exact opposite in my case. I create the same style every time and sometimes the song is too fast for that vibe or style. I try to use the BMP to force it to stay in the standard. But just work sometimes.
I"ve been watching your videos and enjoying your content for a few months now. This video was unreal! I'm making my morning coffee playing it on the TV and my wife comes into the kitchen, she can't see the TV in the living room, and asks... "What play is this from?", then asks why I look so perplexed. I then walk over into the living room and see your face, and we're making the same face. What a time to be alive!
@@acageybeard6478 That’s great!😀
Always a pleasure watching your vids my friend! Let's talk again soon!
Thanks Bob I just subscribed! You are a very admirable person. You don't need to worry about being long winded or playing full songs I think even if your video were 4 hours long I would still watch!
Super kind of you to say!
I use Audacity (free) + Al plugin for it (free) separation voice & music, or 3 Instruments + voice.
It work Very nice and fast.
I absolutely adored your honest reaction to your first creation. I TOTALLY get that feeling.
Love your work, sir! Keep up the great content.
I watch your vids and other AI review vids, and I gotta say: You have the best results of anyone else by far. I think your prompts fit the AI better than others. It's your music background, I bet!
That's funny because I'm always getting grief about my musical choices! :D
Really enjoyed listening (and watching your reactions, Bob 🙂). Delighted you're enjoying v1.5, and kudos to you for your creative crafting and thoughtful commentary!
Do you work on Udio? 👀❤️
@@calvinredfern8480 yes, I'm on the Udio team, and also under the same ID in our Reddit and Discord communities 🙂
@@UdioAdam Hi! Can you give a quick summary of where we are with the law suit junk. I love Udio and generated some good songs but am afraid to do anything with them. I mean are people copyrighting these songs especially when using their own lyrics? Are people putting these songs on CZcams, are they distributing them through Distrokid, CD Baby? Where the hell are we? Ha!
Stems is not the most important update for 1.5. It is Remix. You can change a song completely with the new remix feature. I wish I had your video and speaking ability to do a video but I don't. So I point this out so you can check into the updated remix feature and may consider doing a video just on remix. Believe me you could.
Hi! Can you give an example of how remix works? What changes can you direct it to change? Thanks!!!!
Glad to see these things improving already! Honestly, my only hope was to get the vocal stem from the most recent song I’ve worked on. A 12 year old song that had my own instrumentation but no vocals. I used Suno to get a sweet vocal track for it and I extracted the stem in FL Studio and I’m extremely happy with everything about it other than the quality. Sounds like a chorus sort of effect. So I was hoping I’d download the stem directly from Suno and maybe it would be a massive improvement. But I’m dissatisfied to find that it sounds exactly the same
1.5 is really good but I find it generates a TON of hallucinations. In my experience you have to play with the lyrics a bit or use the trimmer thingy to get the timings right. If the timing of the lyrics is off, it throws a wobbler.
Such a joy and delight to watch your reaction. Gave me goosebumps. Thank you for sharing. I’m inspired! 🙌✨
That would be a nice project... Remix in a DAW using EzDrummer and Ezbass and maybe Scaler to figure out the chords. These stems sound like they're actually useful. Overall sound quality has really improved a lot.
I wish you could update lyrics only on remix. So often I want to tweak a line (like class vs glass) or spell a word phonetically to get a better sound.
I had the same look of disbelief when I created a gospel song and I almost cried listening to it
I watch this because I love your excitement.
Love hearing that. Sometimes I really "dork out."
Incredible. The jazz song is something you should bring into your studio and work on it.
From my experiments on the Udio Stems : 1) are more fast and confortable to get than external services 2) are way better quality than any external service I tryed. All of them had the bad habit to downgrade the bitrate of resulting .wav (my Cool Edit regulary warned me about) Also, with Udio Stem I didnt hear that "hoovering" effect on human voices. So Udio Stems are really a good step forward... BUT in some cases, I also heard a clear voice echo in the "Other" track. This means they are not perfect and NO, unfortunately, in most cases you cant have the "Instrumental" version of your song just mixing "other+bass+drums", because some voices will be still around...
Oh, if you didnt find out yet, you can see the seeds and entire settings for old songs
Nice,but where can I find these?
Where? I have a couple of old ones I'd love to redo,
With all of the AI swirling around I feel like I have joined the land of the luddites with my #HumanMade music. Sigh... But I will admit that Udio has become most impressive. As soon as the ethical issues are sorted and more stem options are available Udio will be an amazing tool. Suno is playing catchup it seems.
Human made will always be special. I mean, we don't use GPT to do our journaling for us, do we? Human made music will always be the most personal and therapeutic. I love AI music, but to me it's like an abstract painting where you throw paints at the canvas and rejoice at the serendipity of what you get: it's fun and amusing, but it will never be as personal and therapeutic as making "natural intelligence" music.
Look at it this way, there are a lot of people out in the world who have always wanted to make music such as myself, but I was never able to learn any instrument very well for whatever reason I don't know. Classic white guy syndrome, no rhythm. Now I am able to go back and get all of the music I wrote years and years ago and bring them to life, with the help of AI.
@@TheAceTroubleshooter That's awesome, and in itself it's more than reason enough for AI music to exist and thrive. I'm just speaking from the perspective of a musician (one who doesn't despise AI, much the opposite of that).
@@VampyressVA I think AI is fantastic. I just hope they can sort out the IP potential issues.
@@spadogs You know, our point of view is the minority, I think. I'm seeing a lot of vitriol towards AI from musicians and artists. I don't think it's warranted, though - like I said, human made art will always be important and special.
I agree, the IP issues need to be sorted out. There are not infinite melodies though, it's a combination of a finite number of notes. As the amount of songs being created each day skyrockets, it's inevitable that many will sound like existing ones... not sure how that's going to be addressed.
Something I love with stems is how you can split out the vocals and then put them through another AI to change the voice. But what would also be great is if you could take an instrument say the guitar part and use AI to turn it into another instrument like piano, or saxophone. But have the melody and timing etc stay exactly the same. That would be incredible and really open up remixing songs.
Well, to some extent, you can do things like that with Kits and Audimee, both of which can convert various sounds into instruments, as well as RipX which gives you some ability to replace the instruments in the stems. None of it is perfect though.
@@BobDoyleMedia Thanks, I'll have to take another look at them.
Great video! Looks to me like Udio is pulling ahead in this game.
I have found it hard of late getting modern sounding tracks out of Udio. Had put it down to them maybe removing a lot of content from its training data.
@@epicchannel4724 Interesting theory. I really don’t know!
Udio has never been great for modern music but it did excel at older music until they changed the model, now its ridiculously difficult to get anything usable out of it on 1.5 or the original 1.0 model. Suno is far better at modern music but sounds too robotoic for me and the audio quality sounds like a bad MP3 rip.
@@maccagrabme That's pretty much the long and short of it. Udio has MUCH better audio quality and natural sounding voices, but is also much less reliable in making good musical choices. 9 out of 10 are completely off the wall and nearly unlistenable. Suno, on the other hand, can crank out very listenable and very enjoyable music almost every time, but the singing voices are very noticeably "robotic", except for the rare exception. If we could just get Udio voices inside Suno I'd probably be very happy.
I haven't used it in a month because of poor quality. It was quite good before.
@@TheWayOfJamesBandI get very good voices in suno. It depends on the genre and extra prompts. The common example is "soulful" RnB, instead of just RnB. I also get good results from rock and emo genres. But it depends. 1 out of maybe 10 generations is quite good. Others can be average to terrible.
Glad to see one of my songs featured on your video
That is mind-blowing.And thank you so much for your efforts in exploring and explaining this AJ stuff that's really not so easy to grasp for guys like me (65+)!
I want someone to come out with an A.I. app that lets you load a song and it accurately spits out the music notation like the chords for the song the melody and more in MIDI form or something similar.
Fadr extracts midi. I tried it a few times. Wasn't too bad.
There have been a few for years already.
wow, great results with version 1.5
You rock, sir.Thank you for your help and quality hard work in getting these videos to us dreamers...
Getting optimal results with Ai stem separation usually involves trying out a few model variations and picking the best one. They may need to diversify by genre to get the best results.
Having used it for 2 days, I’ve found 1.5 sounds better, yes. But I’ve found it a bit harder to prompt fused styles, and off-genre electronic, for some reason. Especially in the 2 minutes outputs.
10:44 Which tool did you use to make the AI video?
About the video... One of my concerns was... It will be a stem separation or a stem download option. I mean. When the aí make the song it make it "mixed" and after remove the vocal (like we do) or it make de song separately and we can download one of this? My opinion Udio have the stems before the final "mix", or the separation system they use are so much better than Suno(based on your examples cause I didn't test any of them). It seems like the Suno voice stem have the same aspects of a separation made on Vocal Remover, Moises or another service like this. I'll test, but what you show gave me an idea what I can expect. 😊
I love seeing Bob's extreme excitement with the music being generated. It's sweet & endearing. 🤭
You are knowledged. Thank you for feedback and reaction. ^_^
Love it, Bob. You could show us how to separate stems in Udio and change the instrumentals like the bass and drums and mix it with your own bass and drums :)
🇧🇷🇧🇷🇧🇷🇧🇷👏🏻, Wow, wow, wow, my word! This is amazing...
Do you publish any of your Udio gens? if not, you should start with that first candyland song; I'm sure that'll trend, it was great. You can't get duets like that consistently with one prompt.
I wonder why no one has created a stems outpaining for stems quality meaning that it would take the dirty digital sounding stems with artifacts and predict how these should should as clean signal devoid of those glitches. It should be super easy for someone who knows what they are doing to have a dataset if true stems vs ai generated stems for the same songs and then create a engine that would provide stem cleaning.
It would be more like an upscaler, if we keep the image metaphor.
@@FrancisBourre yeah I blanked out looking for this word.
I recently psuedo-"upscaled" some noisy suno songs using the Remix function on udio with a low variance setting 0.1 to 0.15. It got rid of almost all the noise and rebuilt the vocals and instrumentation, keeping it very close to the original with some minor adlibs, absolutely nuts. Though this was done on a whole song, I plan to test it further on a noisy vocal stem with "acapella" in the prompt
@@MCG3333 that's mighty interesting. Can you explain the workflow in detail? I'm more of a Suno user but I tried applying a low pass filter to a stem I needed with hopes it will treat the provided sample as a fiktered-out intro to the full melody but it just provided songs that had filter applied throughout.
@@MCG3333 get it now, will try it tonight, thanks!
Incredible!! Thanks for the heads up!
Hi!
this is Edson from Brazil and I really enjoy it.
Hi Bob I tried this function I find it far below average it seems that there is a kind of phaser on the channels or stems that I used to use in the studio on my guitar It is still far from Studio quality
@@rubekker5572 my experience is that results vary. For the most part I get really nice, clean stems, but I think it’s ultimately gonna depend on how complicated the mix is.
Really really enjoyed this. Subscribed
I use Suno... might wanna switch. :)
Their v4 might be out soon
Great channel. I find it very useful and informative about these 2 programs....possibly might make me switch to Udio! Love your enthusiasm! I think Udio seems to give much more realistic vocals than Suno, who often sound ever so slightly vocoder-ish.
I'm in love with this video!
To copy the seed of a song, open it and hit the Create button (next to Share / Edit / Publish) and chose, well, yeah, "Copy Seed #"
Excuse my lack of knowledge: Could you please tell me what this seed of a song is for?
@@andrewmartin8565 The will help you make the same version of the song. So if you liked a song made with udio 1.0, you can select it, put it in as the seed for a creation with udio 1.5, copy all the other settings (style, lyrics...) and you will get a improved version of that same song. Or you want to use other lyrics, but have udio use the same music, you also can use the seed.
@@TheWritingMarmot Oh, nice. Thank you!
I've been using Suno for a while now and are quite happy with it. However, where it shines in some genres, it struggle in others in terms of quality of the different instruments being used.
The sound quality of a flute for example, can in many cases sound terrible because of the poor sound quality, same goes for the female vocals sometimes(or any high pitch sound really). I wish it was possible to choose different vocals from both female/male
You can control the voices by prompting "sopran" or "bariton". I always use "bariton" because the lower voices sound more realistic with less artefacts.
@@andrewmartin8565 Thanks Andrew, much appreciated. 👍
This was great thanks, I use Suno and the leak makes it unusable, Udio sounds amazing
Hi , the video part is also made by Udio or Suno? Congrats for your work . Thanls
Wow fantastic! I was wondering, does the AI also create the videos that match the lyrics, or did you create it later with another tool? if the latter, what tool? Thank you, just subscribed.
Dam were her vibing and dancing to this crazy stuff! awesome. 🔥
I noticed I got a lot of bleed with the vocals and the instrumentals too. Can you tell me: does RipX do a better job of separating the stems?
Actually doing some kind of side-by-side comparison might be a good idea since so many are entering that ring now!
I can SO relate to your expressions on the first gummy bear song! These moments are the best, jaw dropping.
And big companies want to take this away and control AI music themselves...
Suno is better than Udio, both are useful for real musicians, and both suck for UI design. Once your library is large, going back to find early works, well, be sure to tell someone where you’re going first, because you never find your way back. Nice to see this update improves Udio’s vocals, which garble text often.
Whoa! Unbelievable generation of the "Land Of Dancing Candy" song, Made me pull out my old Candy Land board game and dust it off for a few rounds, 😂
Suno is where it's at!!!!
is there any AI software you can put the lyrics in and your own melody! because i would love to hear a choir sing my melody and lyrics. i know there’s band in a box but it won’t do vocals choir thanks
Loving the information you're Putting out there. However, the tasting music makes me want to claw my eyes out. Lol,
thank you for the video, i was curious on how the stem separation on suno competes with the other free Ai stem separation .. seems to still have vocal bleed plus it does not separate tracks.. im gonna hold on subscribing suno for a while more
I love it - I creat songs in diffrent languges - and pronanciation is better now - less mistakes.
It's incredible really 😊
Iblove your reactions because I can tell they were totally authentic. It is obvious you have a very sensitive soul, your non-verbal language always tells the truth. Very touching video Bob! 🫸🏻🫷🏻😎
Wow UDIO is the clear winner of this,, I already tried them stems seperation on suno before i watched this and after just 1 minute of playing around with, i knew it was unusable. I mean considering its AI, i expected high quality vocal stems, not WORSE quality than the vocal seperator. So my biggest question is "Why add a feature that is unusable" Same thing with their new design,, is it just me or was the old design much more,,, you know,,,, good?
But after watching your video, the UDIO song AND stems was so high quality and i think i was just as surprised as you, i mean, just wow... So guess it´s time for me to switch to UDIO untill suno steps up their game.
So thank you very much Bob for sharing this, i enjoy your videos and are always looking forward to some new exiting AI tools to use in my music creation, thank you for being here ;)
Yea Udio always been a couple of steps ahead.
Hi! Crazy stuff, nice job. Do we have any idea where we are in legalities of using this stuff, distributing it. I'm not sure how to approach this without knowing the rules.
It's unfortunate that this video doesn't highlight why so many people are actually unhappy with version 1.5, due to a lot of instruments sounding like MIDI and unnatural vocals. You got a good generation. Not all of us do.
Well thank God you set things straight. We were all thinking it was good.
@@bobhawkey3783 the only 'problem' I've had is: choosing which new clip to use...It's like choosing a favorite child!
That's why you can generate songs untill you get the perfect one
Play with the clarity. I've noticed that if I bump it up to 100%, it indeed starts to sound like a MIDI.
Love your videos, especially this one. I’ve released 20+ songs I’ve made on Udio, so am very excited about this update!
I was excited to see that Suno has stem separation, but they didn’t do it, right…. I can still hear the AI vocals bleeding in the instrumental. I thought they were going to make it to where the AI has his own track and the instrumental has his own track but literally separate with his own program instead of it separating the vocals in the instrumental like every other AI does it if that makes sense…. They should have it to where it automatically does it from the start meaning if we generate a song, it automatically releases two tracks, the a cappella, and the the instrumental without us having to use the AI to strip the vocals from the instrumental 😂 I hope I’m making sense
Totally. that's what I was expecting as well.
I agree, that would’ve been great. But, on the other hand, I think it would also be extremely tricky to pull this off, because they would need to have access to the multitracks or stems, which they almost certainly don’t have in the vast majority of cases.
I agree, that would’ve been great. But, on the other hand, I think it would be extremely tricky to pull this off, because they would have to train an entirely new model, and have access to the actual multitracks or stems, which they almost certainly don’t have in the vast majority of cases.
I was so excited too until i heard the instrumental stems :( :(
@@ivanislasbe Yes, it seems the quality of the stem separation isn’t fantastic unfortunately.
If you take a screenshot of yourself with your average lighting setep, then feed that image to an AI and let it create this type of background for you, but then hopefully the lighting will look more coherent and it will look more realistic, or even just striahgt up cool. Of course next you have to cut yourself out using maybe something like photoshop filler of empty spaces on a pictures. Or maybe another IA can do that to.
I think I asked this before. Not sure if I got reply or not. I notice the videos showing as you were listening to your Candy Land creations. My question is, how do you generate those videos? Are they part of Udio or did you use a 3rd party tool to create the videos? Thanks
I have a challege for you.. With AI (udio,suno) make an intro (or two) Like a DJ Intro or a like a enter the ice intro for hockey. Is it possible, presentation voices, epic music, lots of tags to create the perfect one :) Are you up for that challenge?? (And ofc it make me very happy if you are able to make it good! )
LMAO.... aside from loving the product, I just love your reactions! Bob....
I'm genuinely blown away by this! :D
How do you make this great film song it's so cool!
Love it!
Hey bob what the ai you used to create the video the singers singing
Hi awesome video. How do you generate this kind of video for the song?
cant find a download icon, and when i go into download text , no stems
Hi, I'm trying to create a breakdown for Metalcore in UDIO, but I'm not having success with the prompts I choose. What can be done? Cheers!
HI from Italy. How do the video?
I see video music for every song...how is this?
I prefer Suno,in my opinion make better songs and more catchy,and in less time,but the quality of sound is little poor.
In my opinion Suno must improve in this.
Sub'd. Good job 👍
Where did you get the video at 10:30 from?
Holy xxxx. I just tried. That is crazy. Seriously this will freak pro music production out. Holy xxxx
As the most music AI that I know: A question. My wife sings really well, and has composed several songs. Isn´tthere an AI that the input is a capella song and the output is the song with all the instruments? All AI that make stems has the resouces to train one AI able to do that. If it appears, pls tell us about it. And Thank you so much for you teaching videos.
You can do something like that with Band in a Box, if I understand you. If you want to add music to a vocal or other instruments recording, then yes.
i do not find the option for the stems. When i press download a single file is generated and downloaded
Bob, how do you make those videos for those songs???
What you'r saying in 13:14 is wrong, cuz it splits the Stems after generating the whole song, like many AI or even Logic Pro 11 does. If you be a professional producer you know that this kind of separation can never be used in a professional world, because of lack of quality. UDIO engine makes the song in whole, not like a DAW in different channels or Stems. I know it's gonna do that eventually but I believe it's gonna take a very long time up to that point, cuz their users are not pro so they don't need it, and the Company already make a good profits, so no need to push. BTW these Stems are already processed, so nothing you can do with them when you add them to a project in a DAW.
The candy song isn't teenpop though, so it ignored the prompt imo
For me the Suno stems never separate properly
the separations i've been getting are very poor indeed - unuseable in fact!!!
No bpm information on the stem?
Suno has overly compressed voices (which can be nice _sometinmes_) compared to Udio, but has a way more versatile output. It adds bridged on its own, quite surprising musical ideas to chorusses, that like. Impressive!
However, the stem separation in Suno is not really convincing, when it comes to refine that song 'at home'.
Well, my experience is all based on your videos, yet. I am going to get my feet wet soon.
think uido wins on the stems.. i found it is hit and miss with suno and also its not multi split like suno its just vocals and song.. and you can still hear vocals in the background.. but udio Mmmm sweet
They both make music that sounds ok, silly kinda nursery rhymes are not the same as radio-quality catchy pop songs or rock songs that could be a hit, can they make music that's really good that's where the frustration starts because they can but its a long road stumbling around in the dark, controlling what you get is random where right now we are monkeys pulling a lever hoping for a treat as there is no concise manual on what prompts deliver that is put out by developers, they both need manuals on what prompts do they must know how to talk to their software, imagine software like Photoshop with no manual just a line that says push buttons until you find something you like, oh and pay for every random experiment where 95% will go in the trash..
Im more fan of the music from Suno... BUT!!! That stem separation from Udio 😳
What instrument is that? I think it's a muted trumpet.
Is it just me or think that Suno vocals are great but the audio quality seems really low compared to that of Udio?
I seem to get a variety of results, where sometimes it's the other way around.
Depends on the genre and luck. For me, if audio sounds low or bad on more than one generation, I wait and reload after a while. I feel like they have people working shifts, and each person adjusts the EQ different. 🤷