New AI Generated Music Leak is Insanely Good.
Vložit
- čas přidán 7. 04. 2024
- A mysterious new AI Music Generator is making it's rounds...
▼ Link(s) From Today’s Video:
Suno AI V3: • Suno AI V3 is a Comple...
First Leak: / 1777059367788982389
Deleted Leak: / 1776930684642443400
Cryptonomics compiled leaks: / 1777323606151712970
Udio Website: www.udio.com/
Nyxia AI: / nyxia_ai
► MattVidPro Discord: / discord
► Follow Me on Twitter: / mattvidpro
-------------------------------------------------
▼ Extra Links of Interest:
✩ AI LINKS MASTER LIST: www.futurepedia.io/
✩ General AI Playlist: • General MattVidPro AI ...
✩ AI I use to edit videos: www.descript.com/?lmref=nA4fDg
✩ Instagram: mattvidpro
✩ Tiktok: tiktok.com/@mattvidpro
✩ Second Channel: / @matt_pie
-------------------------------------------------
Thanks for watching Matt Video Productions! I make all sorts of videos here on CZcams! Technology, Tutorials, and Reviews! Enjoy Your stay here, and subscribe!
All Suggestions, Thoughts And Comments Are Greatly Appreciated… Because I Actually Read Them.
-------------------------------------------------
► Business Contact: MattVidProSecond@gmail.com - Věda a technologie
The quality of the singing is outstanding. Suno now has a strong competitor for sure.
Not yet. We don't know how many generations it took to make it. Imho with good prompt Suno generate good music half time, so having two songs generated at once usually one is good
Like Matt said: It's not a competitor yet until the original source reveal themselves.
I find that Suno songs in other languages have better singing and sound less muffled then if you generate English songs.
@@chanpasadopolska LMAO nvidia has all the compute in the world, you think suno will have a chance against Nvidia?
@@MHG796 they are Nvidia? I was not aware of it. But still, Nvidia has rather experience with processing graphics, not sound.
The fact that it's a recording of a recording of a recording and still sounds _that_ good makes we wanna hear stuff directly from the model. Damn.
If Suno can implement stems, kick up the fidelity, then Suno will be in the lead again. If they came out with a SunoStudio, then music producers like me, can embrace it. We want control and have the ability to put our human, artistic touch to it. So, I’m not scared, just impatient 😊
100% - and if you could get your own melodies in there it's a wrap!!
Oh man, that took the vocals to a new level.
Though V4 Suno is around the corner with Inpainting.
I WANT to specify BPM, key and specific instruments! And a NEGATIVE PROMPT!
Hopefully V4 will fix the train wreck that has been V3. I've been forced to use V2 because V3 has been so worse than V2 or V3 Alpha.
@@sixstanger00Not in my experience. V3 has been insane compared to V2, at least in the rock genre.
@@MrPablosek Disagree completely. The genre I work in is alternative rock/90s/grunge.
Numerous songs on my "artist" channel (Urban Decay Union), and they were all done in V2 (except for the most recent one, which was V3 Alpha). V3's clarity is not within light years of those songs.
The vocal clarity in V3 has been atrocious, and it has this annoying habit of forcing "choir" singing; it simply refuses to do solo singers.
No matter what genre (and you can check this with ANY of the top songs appearing on the Explore page), V3 will immediately start layering additional vocal tracks after each bar. Even if you prompt it with "solo singer" or something similar, it blatantly ignores this and forces layered vocals anyway.
By the time it reaches the last bar of a chorus, it's layered on so many vocal tracks that it sounds like a rock concert crowd singing, with the beginning and ending of words being so muffled/blurred that it becomes impossible to distinguish what the lyrics are.
I've also noticed that V3 has a much harder time producing unique/catchy melodies/chords/pitches, often just "reading out" the inputted lyrics over a somewhat "generic" rock track.
I'm not alone, either. There are threads on Reddit discussing this, and the "Bugs" and "Suggestions" channels on Suno's discord have been ablaze with disgruntled users complaining about how shitty V3 is.
If you didn't use V3 Alpha, then I don't expect you to see a difference.
V3 Alpha would produce a satisfactory track after 6 - 10 generations. For a recent song I was working on (the first I attempted in V3), I went through over 90 generations and it never got close.
I agree with choir bias. The vocals were much better in v2 but I can understand why they went choir safe in v3 because v2 would occasionally misfire and autotune of solo vocals made it sound terrible but you have to take the Ls for the wins that V2 gave.
As a musician this is like a roller coaster ride. It’s exciting and scaring the sh** out of me at the same time.
This will empower your creativity to greater heights
As a 3d artist, I feel much the same way about video generators.
@@LinkRammer Me too as a coder, we are so baked
I still like live music and remember just like chess players who can never beat AI, people will still pay for genuine article.
The one good thing that could come out of this is musicians being able to play songs written by AI. There are a lot of good musicians out there that are very hit and miss when it comes to song writing.
AI is always going to be a bit flat when it come to sound quality, there's no way AI can reproduce all the details of a real band playing. For people doing electronic music, they are done though.
100% catchy fun soul music
100% Lovecraftian horror from the abyss
i've seen so many cool unreleased AI stuff just die, that i don't even care anymore unless they have some sort of ETA
Stable Diffusion 3? Sigh
Yeah. Unreleased and Leaks suck.
But perhaps it can at least inspire people by showing what is possible.
And somewhere out there are the people with the knowledge of how to do it. So perhaps they will create again someday.
The point of creating AI tools isn't entirely to use the AI tools, it's also to advance their tech. Google doesn't release a lot of stuff because they know it will be significantly better in a matter of months. I just look at it as a nice benchmark for how far we've advanced
@@xitcix8360 I want to play the game, not be a spectator that reads the post-game scoreboards in the newspaper.
should be an AI dj on a radio station playing 24 hr - AI New Tunes
Imagine an AI music radio premise where all music is only played once, if it was good and you didn’t record it… it’s gone. Would be very interesting with the whole way we’ve got used to things being permanent. Like this would go be like going back to the radio days… only way more extreme!
@@Edbrad that's a neat idea
yeah it'll happen duh but im trying to here some trash ass human made music boys
Competition is good. They will HAVE to make it competitive. Thanks for the stream Matt.
i love that you stay majorly up to date on AI music Matt. keep up the good stuff man. love you and the videos
In my opinion, this is superior to Suno in many ways. I have a musical diploma and I can tell that it has a stronger grasp of chord progressions and song structure. The voices sound more human-like, both in terms of the voice itself and the techniques used (vibrato and voice dynamics). Also, it blend everything together like it's a professionnal song 😍
However, what blew my mind was the ability to convey emotion. When I heard that song at 10:53, I was completely blown away. The voice sounded just like a real person pouring their heart out in a story, and the guitar had a genuine, authentic sound that I've never heard from any AI before (guitar is my main instrument, so I was paying attention 😀). The voice and guitar chords are just perfect, litterally perfect. Even the moment the vocal stops, the guitar don't play something boring, it change the chord and the dynamic 🤯🤯. The fact that it can come so close to replicating a real voice+1 instrument sound makes it even more remarkable. If it can do other strings instruments like violon, double bass, [...], that would be extremly impressive, opening the door to classical music, or any funky combinaison.
It's funny how they could convey emotion, something Text To Speech never managed to do! And one would have thought conveying it while singing well would be harder!
Finally we will get diverse music again like the 80s. Diverse and full of new sounds and beats. Love it
Lmao music historians all agree that the 80s was the duff generation. Take your nonsense to the pub 😂
@@sixfeetwonder_ Obviously don't know wth you're talking about troll. BY FAR the 80s was most diverse and best with 70s second then 90s and 60s. The full truth is unattainable to naivety a wise man once said. God help us smh
@@ListenGRASSHOPPER the 80s really were the best for those who like musical diversity. Even artificial intelligence can produce music with more soul than the garbage produced by humans today.
Diverse music still exists, though.
Latent Space Pop
that's better than the garbage that's so successful these days...
Excellent video, Matt.
Thanks Matt. Love the videos!
Thanks for this. Your content is always cutting edge.
It's good that this competition has come along. As a non-musician, I won't judge whether udio is better than suno, but I love that AIs allows me to create songs for the first time in my life. I've been playing with suno for a few weeks now. I wonder how the recording studios will take to it. Will they try to eliminate this threat through the courts or somehow absorb it, which would probably end up making it impossible for me to play with it too. Maybe a year from now it'll be hidden under a pay button that will create songs exactly to my taste, with no way to influence it.
One musician told me he's contemplating suicide after seeing this. He feels like all his hard work and studying composition was for nothing.
His stupid decision won't change anything. As a translator, I was one of the first victims of AI but I adapted. The musician you mentioned must too.
RIP adapt adapt. How about plumbing?
totally understand im very sorry for your friend :(
its hard to accept that now everyone is better then ever musician ever because of these generators
however on the other hand I truly am very impressed by music humans make and will always seek it out :)
It wasn't for nothing, it was just enough to train the ai😂
As someone who loves making music myself, it's not all for nothing to make your own music. Human music will always have at least some value, and I think humans tend to prefer things with a human touch to them.
I really hope things go ok with your friend. It's easy to fixate on the negatives in life but meaning can be found anywhere, including in music.
this video is short and awesome introduction! Thanks Matt, I definitely want to try this out.
How do we get a referrer for fast access?
Incredible, can't believe Suno was eclipsed so quickly. It's beyond impressive.
No better day to get eclipsed
nothing happened as long as its not available
LOL
Wow. Ya, the main one in question was really good. Definitely want to hear some more genres.
thank you for sharing this it's truly amazing the way ai music is going
What we really need is a Magnific AI for audio. Then we can take Suno, with less than high resolution audio, and upscale it.
Ironically audio AI works on Spectrogram images.
@@cesar4729 so try upscaling it
@@apache937 Not quite. Modifying the appearance of the spectrogram has serious consequences when translating it into an audio signal. Although it sounds better with better pixel definition, a highly specific magnifier would be needed to fulfill the function. It's not like there isn't a way, but nothing remotely as simple as "make it bigger."
You can do it
@@cesar4729 train a model with each song in many bitrates then it will probably work with scale?
I've been using Suno for a few months and getting amazing results. This is amazing too. But the mix still sounds thin and needs output to separate channels for each vocal and instrument track. Then each track can be improved (levels, EQ, compression, reverb etc) and a better final mix.
It will be a formidable competitor. Suno will now have to race against time and release better updates. I feel that 2025 will be the year of musical AIs.
Absolutely fascinating breakdown! The rapid evolution in AI music generation is truly mind-blowing. It's exciting to see how competition in the space is driving innovation and pushing the boundaries of what's possible. The depth and quality of music that these AI models are producing now are staggering. Can't wait to see how this shakes up the music industry further and what Sunno AI's response will be. Keep up the great content!
Great vid as always. I am an older cat in my fifties and I just want to say keep up the good work. Your channel rocks. I love everything AI and your channel is my top favorite right now.
Awesome! Thank you!
WoW ! this is insame !! Dude you nailed it.
I have been producing music for over 20 years, and I've learned many instruments and trained my voice to sing acceptably. Yet it's still would take me days weeks and even months to work on a single song and now with AI tools I could knock out a whole album in a day.
AI music it's actually pretty good and there's still a bit of human touch you need to put into editing and tweaking things to make it just right for your taste. It's pretty fun, although I took a while to embrace it I was scared it would take the soul out of music but it seems it's just another tool. The only concerning thing about itis it's powerful enough to virtually replace everything in the music studio from producer to musician to artist.
Now anyone can pick up a keyboard put in a few prompts and listen to some music they want. This is the future of music. It is here and it keeps getting better.
Thanks for sharing this. I made some songs today and tbh I’m blown away by it. There are a couple of minor niggles. If you enter custom lyrics it can skip sections or mangle the words, but the quality of the vocals and instrumentation is incredible. It’s also great that it offers the option to create instrumentals. What will it be like in six months or a year?
This breakdown of AI music innovation is fascinating! It's amazing how these AI-generated tracks are starting to blur the lines between human and computer-generated music. The depth and quality of these tunes are impressive, especially considering the technology is still evolving. I'm curious to see how these AI music generators will continue to develop and compete with each other, particularly in terms of quality and affordability. Also, the competition between different AI platforms like sunno AI and this mysterious new entrant could really push the boundaries of what's possible in music production. Can't wait to see where this leads!
I have a serious question. Does anyone who generates this music feel a sense of pride? A sense of accomplishment? I feeling of "I made this beautiful song"? This is mine.
Does anyone? Because as a musician thats how my own music makes me feel. Im proud of it. I love it. It genuinely is me.
Just wondering...
It is more of a sense of accomplishment that a PRODUCER gets. A process of getting something good out of AI is very similar to finding talent and providing it with resources and direction to get at the final cover song result. Not to be mistaken, this does require good management skills, a good ear, a background in music and a clear sense of what it is you are trying to achieve.
@@michaelkostukov9669 sure but is it yours?
I definitely don't get the 'this is mine' sense that I do with non-ai creations. For me it's more a sense of wonder and beauty of things I wouldn't have any ability to create on my own.
There's also a bit of disappointment that I can't really share my best gens because the market is flooded with music now and whatever I make gets overlooked now that the limited Alpha is over.
I do.
@@kamay7021 lol will it aint u boi
Great channel!
Hey Matt, you confidently say you can hear it’s AI, how can you be so confident? If you had to mix this with a thousand songs in the same style would you pick this one out of a 1000 songs and still confidently say it’s AI? The first song I’m talking about.
I feel that suno is better, but what I do notice is that the singing is more versatile and in better quality in its competitor, interesting.
The music is about the same as Suno.
It’s the singing that is a level above.
wow, speechless after hearing that audio.... thats crazy!
Amazing. I wonder if it generated the music separately from the vocals and then adds the vocals, that would be a good way to come up with this result.
We'll have to see how Suno V4 will sound. And also, they all demonstrate music with vocals, suno is currently the only one that can also provide good instrumental tracks
I never would think that I will make french rap without knowing french😂 AI music is really new deal
Just wow! This is just amazing!
Stable Audio is up to some new stuff too I can feel it!
Awesome!
Thats insane, I can't believe it. Former mastering engineer here.
What will really be helpful for music will be when things like audio engineering, mixing, mastering are handled by AI within the digital audio workstation itself. Then people who are actually artists will be able to affordably put out professional sounding recordings.
AI background vocals would also be cool
Plot twist: It's their way to announce Suno V4
It’s worthless to me until it can generate 140bpm “screaming and farting” dubstep
The bitrate loss still there! Still sounds good
Around 12 months ago from now, you nade the 1st eleven labs video, and some months after it, we see a huge lot of youtube shorts made with voices from eleven labs.
I feel we are again at a inflection point like that with this video,
Probably some time from now, when this tool goes live, we could see people generating some 10-15 second bits which are groundbreaking, and become mad viral, and I'd be thinking how matt covered this reason for that entire wave so much in advance and i saw all this happening.
Interesting.
same
Imagine how far humans would be already if focus was on cooperation :) Combining forces and talent always wins.
Imagine all albums already being created by AI worldwide! ❤️
As AI slips into the music studio, cranking out bops with the efficiency of a caffeinated DJ, we're glimpsing a future where the only thing a record label needs from a female headliner is a pulse and a pretty face. Why fuss over talent when you've got an AI churning out hits like a musical gumball machine?The real musicians? Those who've bled on their fretboards (like me for example Hello) and cried into microphones? They might just find themselves replaced by algorithms that wouldn't know a life experience from a pizza order.So what's the plan? We take music back to the streets, where the only AI we need is Actual Inspiration. We'll create tunes that smell like the city and feel like the sun on your back. Sure, the suits at the top might not notice-they're too busy swimming in cash to care. I use AI, too-don't get me wrong. It's like having a sidekick when inspiration takes a holiday or when my brain decides to be as cooperative as a cat in a bath. But let's not head down a road where we're all just avatars, with our quirks and creativity squeezed into an algorithm. Even my comment was leveled up by an a. I. So am i human now? Welcome to the real matrix people of the World. lol 😂
it is sad in a lot of ways though now everything thats good sounding is suspect. to AI :(
Once the big companies release true multimodal models, this quality will be considered mediocre at best.
I had a conversation with GPT-4 a few months back, about a boss theme I wanted for a game that I make. The theoretical things it suggested (structure, organs, rhythm, composition, and other very specific details) about the piece, were insane. I later showed the conversation to a friend of mine that finished music school and has a lot of experience in this field, and he was laughing, saying that even after 5 years of experience, he could write something of this high level once every a few tries at best.
If GPT-4 was able to then generate sound based on all the theory it created for that piece, it would make everything else sound bad in comparison, and GPT-5 will be able to do just that.
Specific models like these (including text-to-img and text-to-speech) don't have more than 6-10 months until they become irrelevant. They don't have the intelligence to be more detailed or make specific things or changes.
What kind of a prompt got those dofferent layers and styles of voices in the same song? Wow amazing
One big question I have as a producer is: is the output going to lock in sync with a DAW? Is the tempo perfect in that sense? In order to integrate AI music output with non-AI elements, you have to be able to lay the AI stems into a DAW and have them be in perfect time with the DAW's tempo clock.
This is so great. I love this new Music AI technology.
So how do we know it's actually ai, or hasn't been done by suno if the source is anonymous?
I thought the same, cause why they didnt say the source? very weird... 🤔
could indeed just be try out generations by the Suno team and these are cherrypicked results for a 3.2 or 3.5 alpha that is trained internally.
because the company didn't want the leaks its like gta 6 they want the big reveal
There’s couple feature that makes sunauto better than suno . 1. Generate standup comedy. 2. Separate tracks 3. Partial edit
12:23 can't stop listening, damn, this is so good. 🤩
Udio is released!!!
Urgent Udio VS Suno required
¿Can we built a song using both i'm different fragments or is there a compatibility issue?
Now imagine this baked into a humanoid bot, just ask it "play me a song about ..." and it just generates it in a second and plays a studio quality song about your topic of choice!
The future is now
Sounds good, and it will be exciting to get some competition going driving things forward. I'm not sure it's definitively better than suno based on those examples though quite yet.
I feel like Suno is about a year behind image generation in terms of A.I. development in general. It's really decent, but there's still some artifacts and hallucinations. I'd like to see key changes incorporated if that's at all possible, IDK.
@MattVidPro
MUSIC is the KEY to A.G.I.
I'll just be like Karoly and say : What a time to be alive!
I'm curious what they mean by nerfed Suno... Was it better? What changed? I'm a fairly new user.
I was looking into this, apparently people are saying the voices are more robotic now? I haven't noticed myself.
@@justinwhite2725 Ok, I think I can see what they are saying, now. The songs in this video had very distinct vocals and on Suno they tend to have an almost choir like sound to mask some vocals.
the new jimmy song from Udio "feel the Agi" made me fallenable from the chair
It sounds so good, wow, that is very impressive.
What if it’s the Suno team releasing a variant of their own model as a new service? That would be a master move in terms of advertisements and extra income 😉
👋
I signed up earlier today with Udio.
I don't know if the quality is necessarily better than Suno, I think Sudo can already generate a perfectly convincing pop song but it's inconsistent and the complexity and structure is limited. If these are just average generations and it's capable of more then I think this could be the best generator out there but people saying it's 2x better let alone 10x? I'm skeptical.
we need our tap dancing shoes on to move to whatever application wins
It's impressive, but so is EMO which seems like it'll never release. I'm firmly in the "I'll believe it when I use it" mindset with all these AI announcements.
The first leak you shown us where the song it's referencing suno AI, it just feels way more real than suno, the voices don't sound that they have been remastered to a point were they could hurt my ears but not only that, when the chorus of the song comes the AI manages to keep the lead singer singing the main melody while generating the backing vocals creating a harmony of multiple voices, thats something that suno ai really struggles with
Wow! This is pretty epic for the A.I. Music Generation. The singing is ON POINT!
This is fire
1:24 instant regret
I'd say that's fake.
Music sounds AI generated, vocals sound genuine.
It's not released yet so even if the quality is amazing we need to be patient
FYI I think the sun is creating lens flare on your camera. Was a little distracting to me. Look towards your right temple. @ 13:55 mark
Nyxia seems like the clear winner here, I can't believe you mentioned it and then just moved on. Seems like the best of everything mentioned here, I had not heard of it.
The songs are kind of impressive but also flat but good start points and musicians can push it beyond it
I really would love a AI which makes new albums in the style of dead artists which can make them sing again ie. Tupac, Biggie.
Check out those Japanese idols singing an dancing! ❤
I have made so much crazy good music on Suno. Its insane. Im making a whole album with each song is a different scam. It’s fun.
That 90’s Boom Bap actually sounds not messed up at all. I took it as a Hip Hop Group that has two members. Back then there was nuances between members on how they rapped. Usually one is Energetic other is Mellow. I just that it was another rapper with a mellow flow.
no sudden surprise Suno V3 Choir?? Count me In.
these will be so good when they fix the breathyness of the vocals
Wow, never heard some naturally music that created by AI.
Impressive. Not that I love the song... but I appreaciate the depth and reality of it. Watching close on this one. BRING IT!!!
Heres me hoping Ai automates every job and career
🎵 Uud... uud... Udio!! 😂
Seems like A.I likes the words rhythm and night. As a Suno user I never saw anything to stop making me use Suno as of yet. They should work on adding mastering uploads so we don't need to master elsewhere and should enable you to stitch any creation versions together to make a full song in any order, all you need to do is press link, order, or rearrange the order yourself and everything is blended in.
Yes, Matt is an AI in this video 😂
I can't say anything from the clip and surely not that it's better than Suno. I create absolutely awesome music with it. The only thing is that if continueing from a clip and extending it, the quality get's worse, but I am sure they will fix that soon. other than that, I can't complain about Suno.
Jimmy apples says this is 10x better, Ans btw, the source is latent space podcast, they also interviewed Mike Shulman a month ago, check them out.
Fidelity wise some are moving towards being OK at some point but the prompt adherence is all fairly useless right now compared to images in Stable Diffusion.
I got Suno making some really good stuff very similar to that. Definitely can't tell a few of my songs are AI, but I did spend like $26 on just a few songs with tons of trial and error.
WOW!!
Tell us about some updates on Stable Diffusion 3
I just want one that lives right in the DAW and takes the MIDI tracks and guide vocals.