Suno is now a professional tool and it is displacing all current technologies. I am hoping that Steinberg will be able to integrate with Suno or AI tools to allow faster workflow between them. This will be a do or die as a lot of their current business model is going to be heavily impacted by the advent of AI.
I believe that whoever wins the AI integration race will take the lions share of the DAW market. Selling VST instruments and plugins is not as essential as INTEGRATION WITH AI TOOLS.
Things like Stem separation could and should be a big focus for Steinberg. Steinberg has better stem separation that Suno fro example, but seems to limit how many times you can use stem separation on a track)
WHAT DO YOU THINK? (about Cubase integration with AI)
WHAT TYPES OF WORKFLOW CAN YOU SEE WITH AI?
WHAT DO YOU WANT NOW? (In terms of AI integration and workflow)
Please do not turn this into a philosophical debate over the evils and ills of AI. Create your own forum piece if that is what you want to do.
Greg Ondo said multiple times that Steinberg is not planning to integrate Suno and alikes into Cubendo.
Personally, I think that this is a very wise decision for a variety of reasons.
Integrating Cubase with a predatory platform that was made on purpose to put me (and others) out of work? Yeaaah sure. I couldnāt dream of a better āpartnershipā
Cubase, wavelab and spectralayers seems to have really good stem separation. Iām not sure if youāve tried it out lately, but I donāt think they are behind on that.
The desire for more AI always comes with that subtle threat, masked as business advice. āItās marching forward, you donāt want to be left behind do you?ā Hahaha. The grammar makes it sound nice but itās a full on threat. The real statement is, āCapitulate and do as we say or we will come and destroy your livelihood too. Like weāve taken it from a lot of other innocent people already.ā
Generative AI is here today and gone tommorrow. You have assumed that big tech will continue letting you use it for cheap. Not realizing they are currently using the people for beta testing. Then they will take back their tech when they donāt need your services anymore.
If I were to say what I think I would get banned.
Butā¦
AI as a glorified algorithm is one thing. EG PaintShopPro has an AI resizing algo which is pretty good, and I use it a lot.
But beyond thatā¦
Well, I started in the 1970s, with a 3-inch open reel portable tape recorder. Got an Akai DS 4000, eventually bought a Tascam portastudio, along with some outboard fx and a mixer for the 3 synths I was using. Then the 90s came along, and Cubase Score.
Part of the fun was figuring out how to do something, whether it be fx routing for that āuniqueā effect, or just getting a nice mix.
(I used this setup until 2010, (and Cubase Score 2!), when my portastudio went belly up.)
That continues now. Cubase has all the tools I will ever need, and part of the fun, still, is in figuring out how to use them. Like mixing guitar and backwards guitar to make it sound like one guitar.
AI would kill that stone dead.
I may be deaf and incompetent but I still want to mix my stuff myself.
So, no thank you.
Hi ASUNDER Thanks for your reply. I also have reservations about AI. But it is a good musical tool. I have been feeding my original songs into it and the results are stunning. And still sounds like my music. So it changed my mind about it. I do understand the bitterness toward it. Just like AI has devastated the video and graphics arts industries I believe that it will cut a slice thru the music industry. And I donāt like that idea at all either. But the march of progress takes no prisoners. I must have 5 or 6 Digital Interfaces that are no longer supported. The problem with tech is that it redundant very quickly. AI in Music and Art is here to stay. The business side of Ai is headed for a disaster but music and art and video are where AI excels and it is such a narrow subset that i believe it will survive no matter what. But that is just my opinion.
Hi Realist. Thanks for your reply. I have been putting my original songs into it and I am getting fantastic results and it is still my music that comes out, just with better production than I can do. It is fast, sounds great. I want to move it back to my original DAW or Cubase files and then finish the vocals and leads and arrangement in Cubase.
I normally have not had a need for stems as I play all my instruments, allthough drums come from Cubase. The Suno stems sound awful but I think I will use Cubase to do the stem splitting back in Cubase.
So I have a good case for using both Suno and Cubase. It would be handy for me to see the integration. I think many users in the future will see the benefits of using Ai as a tool.
It is really not that different from using VST drum tracks. I donāt do drums because VSTs do a better job.
That is entirely understandable point of view Googly. But I want to use as a tool and it is now powerful enough to work on original music piece and keep the character of the original music. I also started with a reel to reel and went thru the portastudio path. I donāt want it to write my songs just help with the tedious production and it is fantastic for that and also to enhancing my musical ideas.
I was the biggest Atari ST computer dealer in Australia years ago and I remember musicians hating the Atari, claiming it was destroying their ART. Of course it enabled their art and AI as a tool is in the same basket. Just a pity that many see it as high fakery because people do not need to play an instrument at all to make musicā¦and that irks me as well.
Cheers man and it is good that you are still enjoying making music. Music is for the soul.
Hi Reco. I did not see that episode. Did Greg say why they wouldnāt do it? You will see what I think about it in other comment on this topic, so i wonāt repeat it again. But I was also rabidly anti Ai for music and I fully understand why it is hated..Cheating..intellectual theft etc.
But I am genuinely interested in hearing your variety of reasons.
Hi Batteangle I fully understand that point of view. I do not make money out of music but If I did I would also be very very annoyed. Making music has cost me years of my time and tens of thousands of dollars over the last 45 years. Man I can see why musicians are utterly disgusted with it.
I hope it does not put live musicians out of work. But it is going to effect musical instrument sales and Software sales of DAWs and VSTs and some expensive Studios. Just like digital home systems smashed a few of the big expensive recording studios. It is the inevitable march of technology for better or worse.
I treat it as another tool..like a drum machine or a VST. I write my own songs. But the fact that millions Ai songs are produced daily it is just gong to swamp the market with AI slop.
I ask myself .. If canāt stop it so how can I use it? But like I said I do not make money from music.
For me, having things like SUNO in Cubase would make me feel like Iām sitting next to a predator
I use āpseudo-AIā tools for noise reduction, restoration etc where manual labour is sometimes even impossible. I was getting so bad field recordings (speech) that they sounded like people were trying to make ad bad recording as possible. In the past it could take me a few days to bring such stuff to acceptable levels. Now, sometimes it can be fixed even in real time. For such tools, AI is amazing and this is the type of AI, that is improving our workflow and we should get instead of slop generators.
About SUNO: you may not be able to fight it (this is understandable), but every time youāre using it, youāre feeding it information. Youāre helping with development and expansion. Also, when youāre uploading your own recordings, youāre giving it away for free to those corporations, which are made by people who want to get rid of all of us. I dodnāt went through the full legal agreement but apparently theyāre copyrighting anything you upload there.
This is why I prefer AI tools that are not predatory at least. Ideally, made by smaller devs that are too small to dream about global domination and the new world order like those corpo millionaires behind giant AI.
Ps. Yamaha is financing now AI startups. Who knows what it will bring to Cubase in the future. Probably plenty of useless experimental stuff or software created for influencers etc, but who knows. I would be happy to have Macros/Logical Editor that are AI-powered with natural language instead of what it is now.
Hi,
Greg did not provide any further reasons if Iām not mistaken.
Iād rather not be going into further details why I despise Suno and their abominable attitude and disrespect towards copyright , music, and human culture in general.
FWIW, Sunoās CEO reduces music to a āmeaningful consumption experienceā. I strongly disagree.
You have reservations but have changed your mind because you like its output. Thatās a confession if Iāve ever seen one. A bribe works if the price is right, and every man has his price.
Every time Iāve seen someone grifting about why AI is good (or anything else thatās obviously poison), itās because it serves them. And so when talking about it on forums they are generating a moral defense for their own conscience sake. They know they are participating in evil, but they want what they want, so they bounce moral ideas off the internet to see if any justifications come back. Because if everyone is doing it, whatās the harm, when itās a victimless crime, right?
The history lesson is always bad logic. The telephone, the lightbulb, the motor vehicle; these things help people, not hurt them. AI is not the same. And the amish have the highest happiness stats, while the rich have some of the highest suicide rates.
ā¦
The discourse AI bot just helped me with the forum tutorial. Perplexity helps me with windows issues, when forum devs donāt respond. AI has itās place. But that place isnāt making art, I donāt think. Iām not against AI, Iām against generative AI. Art belongs to us, not to bots.
I just saw a video about the store that was entirely made and managed by an AI bot. Youāll never guess what books it chose to stock on its shelves. You canāt make this stuff up.
We all know these things instinctively, but every man has his price. The opposite of love isnāt hate; itās narcissism. If something serves him well enough he compromises, turns and betrays his own kind.
ā¦ and Eve saw that the fruit was pleasing to the eyes, and good to make one wiseā¦
The notion that AIās place isnāt making art is an interesting philosophical stance, and one I (at least mostly) agree with, but there is also an underlying question as to what constitutes āartā for any given individual, alongside how that individual might, or might not, choose to use generative AI.
Though I have had some peripheral involvement (more on that below) in some music-related generative AI projects, my primary use of generative AI has been for creating cover art for my albums and singles. Sometimes that is just using the generative AI features available in Photoshop to remove something or add something in work Iāve created myself or started building with licensed stock photos or artwork to that point. Other times, though, Iāve come up with an artistic concept and used the generative AI to try and realize what I have in my head but that I wouldnāt be able to implement based on my limited āartistic skillsā. I quote āartistic skillsā because Iām talking about things like my drawing abilities (especially with a mouse), my understanding of color blending/shading/etc., and so on. In some senses, these could be considered technical skills, or a combination of technical skills and knowledge.
In all cases, though, Iām also doing a lot of additional creative work to get to and/or from the point where the generative AI is used to the ultimate result. So what is āmaking artā here? What if I were a talented sketcher but needed help in turning my sketches into full-color artwork? Does my bringing in an assistant to colorize the work, under my direction (i.e. since Iād know what I want but just wouldnāt have the actual skills to achieve it), negate what I did do as art? And what about the assistantās contribution? Does that qualify as art? Now what if the assistant were generative AI? Or what if it were the other way around, where I told generative AI what to sketch, then I did the work to bring that sketch to full-color life?
But back to music and Sunoā¦
I have not used Suno myself, but I have been exposed to it through both some experiments by some of my collaborators, and by my own work in trying to help one of my collaborators work around Suno limitations with respect to getting to an acceptable recording of one of our songs.
One thing I can say pretty definitively at this point is that Iāve been very unimpressed by what Iāve heard from the songwriting side of Suno, and that is something I could never imagine myself using. For example, with my permission, one of my collaborators took one of my lyrics (from an existing song that already had music and was released) and had Suno come up with a melody and recording. He had it do two generations each in four different genres. Most of those ended up sounding (i.e. recording-wise) reasonably good, but with clear issues, for example in breaking up lyrical phrases unnaturally when it came to the music. However, there were a couple that were better, and one in particular had a result that, if Iād sent a raw lyric to a musical collaborator, and that collaborator came up with that result, Iād definitely have pursued the collaboration further, maybe asking for a few minor changes.
But, from every single one of the whole song writing examples Iāve heard from people, Sunoās lyrics have been pretty lame. They often have interesting images in them, but poor lyrical skills, for example with a lot of them having singsong couplet rhymes ā the greeting card poetry-type thing.
In another case, one of my songwriting collaborators had started using Suno in order to make recordings of songs heād been involved in writing over the years. He is mainly (solely?) a lyricist, and not a singer or musician, and I have to say I was very impressed by the first recording he put out this way. This was Suno only doing the recording of a fully-written song. He approached me about possibly doing the same sort of thing with some of the songs weāve written together. (FWIW, Iāve already released some of those songs with my own recordings of them.) Despite some misgivings about the whole AI artist thing, and needing to figure out how to handle the administrative side of things for this, I agreed to go along.
There was one specific song of ours that Iād always really liked, but which Iād never released myself because it is a female song (a mother-daughter thing) that really doesnāt have a good way to make it work for a male singer. My cowriter took the piano/vocal demo Iād done back in the late 90s and used that to āteachā Suno the song for purposes of generating the recording.
My first impression of the initial recording was that it was pretty amazing in terms of doing justice to the emotion of the song, but there were issues, where it got things wrong here and there. So my cowriter went through multiple iterations, trying to get to something that would address the issues. Each generation might improve some things but make others worse. This is the kind of thing that, if working with humans, youād request changes, and theyād just make those changes, keeping what was already working there. Not with Suno, at least not with the Pro plan he had (I donāt know if Suno Studio would have made a difference on this front).
Ultimately, after some frustration with a bunch of iterations, I suggested he have Suno generate stems from multiple versions, then Iād give a shot at comping one recording from those stems. They at least were all at the same tempo, and the āsingerā appeared to be the same across them. The arrangements sometimes had different structures, though. I found was Sunoās stem separation was quite iffy (e.g. putting some things in the wrong stems), so I ended up needing to just select the best version of the instrument tracks and aligning all the lead vocals to that structure for comping. It mostly ended up being some detailed comping (including having to split phrases or even words between different versions occasionally), and there was one part near the end of the song where none of the versions quite got it right, so there was some tweaking of the closest version with Melodyne to get that to have the right notes. Luckily, the background vocals stem was āclose enoughā to work because that was one case where it would not have been practical to comp them from different versions due to the stem separation issues. But I will say that I was very pleased with the end result.
Getting back to the art, though, our obvious āartā in this was the songwriting. The Suno part was the recording, with some minor degree of human guidance from my cowriter. My work on the comping was really pure production craft, though I suppose there was a degree of art in terms of some of the comping decisions.
In my general songs and recordings, my art tends to be the songwriting, the arrangements, the playing, etc. But there is also a lot of craft, and there are certainly areas where I use technology to help with that craft. And I could definitely see using AI assistance in areas where it might be possible to improve my results by doing so, in the same sense that virtual guitar instruments over the years have progressively helped me improve the āguitarsā on my recordings through a combination of responding more naturally to playing the parts on a keyboard and/or programming capabilities. The recording I just finished uses lap steel and fiddle, with the Indiginus instruments. They respond pretty well to playing, but I ended up also doing a lot of keyswitch and controller programming to make them more realistic. If I had some kind of AI that would take what I played and help me avoid the programming side to achieve even better results, probably way more quickly, I donāt see that as demeaning the art aspect of my coming up with the part. (The one generative AI thing I have used in my most recent two recordings is IKās ReSing, where it takes a part Iāve sung and changes the voice. Itās way more efficient using that than trying to do a similar thing for layering my background vocals with other voices using Omnivocal.)
I guess what Iām getting at is we all need to distinguish what, for us, is the āartā part of what we do and what is the ācraftā, and where weād prefer to spend our time, and how best to achieve the results we envision. From my experience in using generative AI for cover artwork, I can definitely say that it allows me to produce better results, and in shorter timeframes, than my artistic technical skills would allow, and, for me, the āartā is the combination of coming up with and implementing my concept. I see a similar potential for it in producing my music. I tend to be pretty slow on the production front, and a lot of that slowness tends to be related to trying to get better results from virtual instruments that donāt achieve natural results from just playing them on a keyboard. Thus, Iād be very happy to have generative AI be able to bridge that gap.
As this threadās original question on Suno, I canāt say any real Suno integration interests me to any degree. I suspect that, if you have the high-end plan with Suno that includes Suno Studio, you can already get a fair bit of ability to move things into Cubase for further work, but Suno itself doesnāt interest me to any degree (unless perhaps if someday I get back to just wanting song demos quickly ā then maybeā¦).
Hi Rick Paul That was a good read. You really hit on some good points in that essay. The fact that AI gets parts of the song right in different sections with different generations requiring all that comping.
Tools that can help with the importing and aligning different takes is an area where I could see Steinberg easing the workload. They already have specific tools developed for that but not really focused on importing and comping multiple stem imports.
Most Steinberg users would, I assume, write their own lyrics. So I will avoid that for now But the ability to import multiple stem takes of instruments and align them is a biggie for me. It is tedious and slow.
I see most musicians will eventually come to the realisation that it is a tool and will want to use that tool by moving phrases of music to and from Suno or other AI platforms. In a generative creative way.
You are correct when you stated that Suno stems are just awful. Steinbergs stem splitting has far higher quality from a stereo wav file.
I found ONE GLARING PROBLEM with Cubase and splitting stems. It seems to not allow it more than once and the item grays out after you have split the stems once and is then unavailable.
Where do you publish your music? I would like to hear it.
Hi again ASUNDER. That is a rather humorous post on the AI shop. The march of technology takes no prisoners. I am an older man and donāt have a lot of time left. I just do not have the time to fight tech and be a purist. I donāt like the way phones spy on me. But I need a phone. I donāt like the noise of a 2 stroke mower but i have to mow the lawn.
It is a tool. And it generates a lot of passable slop. But in the hands of an artist it can help to produce gold.
The fact that it simplifies all the tech production to get a good sound is a boon to me. I can, and do, write great songs but the engineering is my downfall. I have great songs that sounds like rubbish. I am a home musician and essentially i just play with myself. It gives me a human like band to play with. I am never gonna learn to play a violin, or a cello or a french horn.
I really do understand why you hate it. I donāt have the time to change human nature either. It is what it is.
This is an area where I do wonder if Suno Studio might be able to do better than just the whole recording that my collaborator was doing it. At this point, Iām not sufficiently interested in Suno to even try it out myself. But I do like to keep reasonably abreast of whatās going on out there.
Iām not sure something meant specifically for stems would be all that useful, but, if there were easier ways to import different parts of different arrangements ā i.e. not specific to use with generative AI ā I imagine that could be useful. In the case I was referring to with Suno, one version had a guitar solo and another had a piano solo, and they were different lengths. That meant that the vocals being used after the solo in one wouldnāt be usable in the same spot in the other arrangement. Of course, it was a (relatively) simple matter to manually move them where they needed to be. But this isnāt specific to AI. Remixes in general could benefit from anything that made it easier, with one key being to allow for pickups to sections and ring-outs after sections.
I donāt know, one way or the other, whether thatās a fair assumption. I know some songwriters who only write music and some who only write lyrics. I happen to write both (most of my collaborations have been writing music to othersā lyrics, but Iāve also gone in the other direction a few times, including on my most recent song, where I was writing an English lyric to a French composerās existing music ā I then recorded the song, doing all the production myself).
Perhaps, depending on how they use the AI platforms. Iām not so interested in that myself. Iād rather be āgeneratingā (mostly by playing, perhaps with a bit of programming) my tracks myself but then ask some AI capability to improve on what Iāve done. A classic example for me might be my playing a MIDI fiddle track, then wanting a tool to make it more like a real fiddle playerās performance, keeping the notes and basic timing, but doing the āfiddle thingsā that make it feel more like a real fiddle player, without my having to spend a bunch of time figuring out all the little nuances and figuring out how best to program them with keyswitches, CCs, and/or whatever. But I know different people will have different wants and needs, and definition of what their particular āartā is in any given context.
Iāve only tried the new Cubase stems splitting once, and for a songwriting project where I was working with a French language song and only wanting to split out the instruments from the vocal so I could write English lyrics. After tracking the melody in MIDI, I used Omnivocal to add English lyrics as I wrote them so I could send work demos to my collaborator from time to time. I have used the SpectraLayers Pro stem extraction a little more. It has a lot of flexibility, but it still gets things wrong in terms of what parts go into which stems, in some areas of any attempts Iāve made. I was thinking Suno should have been better in that it is making up all the parts, so, ācouldnāt it just generate each part as its own stem, rather than needing to āunmixā the song?ā Thatās what really surprised me about how bad their stems were.
Iām on all the music streaming sites plus Bandcamp. My website (https://rickpaulmusic.com) has links to my music on various sites in the pages related to the albums and singles (thereās also a āquick linksā page with links to my artist profile on the major sites at Quick Links ā¢ Rick Paul ). FWIW, the current top two videos on the home page had some AI involvement. āReaching for the Starsā used ReSing to double my background vocals, and the Lyric Creek Music recording of āSheās Got Her Heartā is the one that had used Suno, along with my comping in Cubase, for the recording.
If you donāt know what music is, that doesnāt mean nobody knows. If itās a mystery to you, that sounds like a you problem.
Or is it that you used to know but are back to the drawing board because you like this little AI toy you have, but it transgresses everything you know about music. So instead of acknowledge that youāre on the wrong team, just redefine reality for yourself. Everything is acceptable when reality is whatever we want it to be.
Because we all know AI stole from all the artists and is ruining their lives and you couldnāt care less.
You donāt know what making art is? You donāt know because you donāt care to know. Because as you expressed here, you care about the result. The people calling AI a tool because itās a means to their end (it functions as a tool to them, by definition). Their end was never making the world better; no, their end is using others as a commodity, to end up as ābetterā than them, for ego sake (by the measurement of comparing success). The reason they canāt make a song thatās worth a damn (or understand what music is) is the same reason they defend AI. Because they couldnāt care less about other people.
As I said previously, AI is disclosing who is real and who is fake.
You have buried the essence of the issue under a pile of stories.
I too help myself to graphic images; usually memes. But Iām not a graphic artist. I draw the line at music because I would rather get nowhere than betray my musician brothers. Not that AI will get me somewhere, because AI music isnāt music at all.
You guys donāt want to be purists? Ya I can tell.
When the sheep gets stuck in the mud they stand there and cry for help. When the pig gets stuck in the mud they wallow in it.
like those corpo millionaires behind giant AI.
