as Suno DAW seems blasting (i dont know it personally)
Thinking of it, maybe its not a bad idea to go AI path to generate tracks,
lets say i want an acoustic country style strumming guitar, or a funky elec guitar rhythm chops, anyway for some production that we don’t use real guitar session player , ill use loops, or more likely guitar vsti, and it sometime lot of programing, finding the right patterns, sound, or build/play it yourself and its not that convincing anyway.(unless working very hard for the nuances)
in this situation, i think AI generated could be nice alternative for quick track built, following track chord, and enable to modify it for your taste.
i mean AI DAW in this sense, can save lot of times, and it can offer so much more, like converting human humming to trumpet etc
but in my view, using Vsti in the guitar example i suggested, its kind of “fake” guitar anyway to me, i mean dropping pre made loops, or imitating guitar by midi chords with auto strumming !
its not a real guitar playing anyway, so thinking about it, if AI makes the same or better job for generic strumming guitar for the fraction of the time and with producers/user chords and ideas. who cares, its another option in the tool arsenal.
its not about making a all song by text or something similar
I also vote no. Or if yes, please make alternative Cubase version for this.
Generative AI is a misery and a lot of musicians is trying to be as far away as possible from it, for their mental sake. Having it in our favorite DAW is not gonna help our (collapsing with every year) mental health.
AI used for creative purposes is one thing. Like advanced noise reduction.
Generative AI that is made to replace musicians to satisfy billionaires ego and wallets, is another. Making “loops” with AI is just another step to the final goal and to normalize musicians replacement. Musicians who can’t see it are simply blind.
Now this AI debate is finally getting at least interesting. Instead of the typical kneejerk negative reactions, it’s really a much finer line than most broadstroke arguments consist of. Sure, AI taking over jobs isn’t a pleasant thing at all. But we all know at this stage, that AI is whatever humans train it to be. From a producer’s standpoint, I’d rather AI be able to help make samples alot more realistic, than just sit there all day comparing manually sampled libraries for their “realism”. The usages of AI is a very multi-layered issue, so I’m actually looking forward to those who can see the usefulness of it for their everyday tasks, rather than the typical broadstroke “AI isn’t the human touch” sort of perogative.
if it was depend on me, i would be happy to go to tape or only audio recording era, no singer karaoke performances , no too much DJ, all musicians would work day and night in studios and on stage !
but the realty is elsewhere, even older DAW replace musicians more than 40 years. well its not AI but using midi, loops, or whatever to mimic real instruments or musicians its not new only to AI.
AI DAW like Suno is a next step to it, and it bites also musicians producers singers and studio shares.
but again, in my view especially as a keyboard player with project studio, i don’t see why i would prefer doing complex midi recording and editing to mimic guitar if i know that i will not record a real guitar anyway in certain project, and AI will generate it much quicker and possibly better
You think? I don’t think so. Staying with your example the creative choice is to have an acoustic guitar strumming in the project. Creating the track, getting a guitar sound and creating the basic strums is busy work, which I call “administrative work” (maybe not the best term). I wouldn’t mind that a software based assistant takes care of that.
While I don’t see this as a feature of Cubase itself – it seems to me it would more likely come in plugin form – this is pretty much exactly where I stand on the potential use of AI in this context.
For perspective, I am also a keyboard player (and singer and songwriter and …) with a project studio. I am decidedly not a guitar player, strings player, brass player, etc. Over the decades, though, I’ve played and/or programmed all the instrumental parts on all except an extremely small number of my recordings. It is something I do partly due to financial limitations and partly because of creative control considerations. I can remember a few times when others have volunteered to play guitar on some of my recordings, and I took them up on it, but I wasn’t happy with the results. It’s not that they weren’t solid on their instruments, it just didn’t fit my vision.
Guitar has been a special challenge because many of my songs are either guitar-centric (despite my being primarily a piano player) or at least feature guitars as important elements. I can generally play finger-picked parts quite believably on a keyboard (I’ve even gotten compliments on the excellent guitars from reviewers on some), especially with modern virtual instruments. Electric guitar leads can be a bit tougher, and more programming is typically needed to add some of the nuances. But strummed parts are the harder thing, especially on the getting precisely what I want side of things. I’ve used a combination of loop-based virtual instruments that do chord detection (my first on that front was Steinberg’s Virtual Guitarist – these days it’s mostly either UJAM Virtual Guitarist or NI Session Guitarist products), but there’s a lot of work involved in finding something close to what I want, and doing any manipulations to get it closer still.
If there were something out there that could get me a lot closer a lot more quickly, I’d definitely be interested. I think my vision on that on the strumming side would be along the lines of I let it know rhythms I’m looking for, along with the chord changes, and it would give options that make believable guitar parts out of them based on its LLM equivalent that had knowledge of what guitar players do (which is quite different from what programming strums makes easy). For leads, it might be that I’d record a MIDI part, but it would add the nuances that relate to a style I described.
Oh, and while I’ve been talking about guitar parts here, I’d probably even be more interested in something that did the equivalent for pedal steel and fiddle because those are two instruments I’d like to use more but have never been able to get sufficiently believable. For example, with the pedal steel (I’ve had the Wavelore Pedal Steel since it was a Gigasampler instrument, though I now use it in Kontakt), there is just a huge amount of programming needed to get truly believable parts, and you actually have to be pretty knowledgeable about pedal steel playing techniques to know how you should be programming it. What I’d like to do is just play a part on keyboards, and have it get mapped to believable pedal steel techniques. Personally, I’m less interested in wind instruments because I’ve been able to do those much more easily with a combination of really solid virtual instruments and a wind controller (I played clarinet and saxophone in my teens and early twenties, but my tone on those, especially saxophone, is way better using a wind controller and virtual instruments).
Again, though, I really don’t see this as an integral feature of Cubase. Maybe it could be one of the “extras”, the way HALion Sonic and various included libraries are, though I’d expect it to be more like teaser level products if something of this sort were included in Cubase as something more specialized would be warranted for more depth and breadth. I think most of what would be needed from Cubase may already be there, for example in the ability to use the chord track to generate MIDI for instrument tracks, general MIDI recording/programming capabilities, etc. I’d see the AI aspects coming in at a specific virtual instrument level, be it for guitar, strings, singers, or whatever. That’s where you’d describe any specific aspects of what you’re looking for to feed off the MIDI (and/or perhaps audio, for example in the case of virtual singers or humming a bassline or whatever).
All I calculate is that if Cubendo consisted of various ai analysis components….built-in Spectralayers Pro (as some clamor for) and 1/3 of the ongoing features-list requests…it would be such a completely out-of-control conglomeration of underlying code that things would be a complete nightmare for the code-keepers at sb!
I like the idea of letting ai make the accompaniment parts so I can concentrate in the melodies. I also cannot play well. But imo the ai should stay away from the melody making.
