Survey on the use of virtual drums

[dgoldwag]

Please take the survey below. We are trying to guage interest in the developement of an on-line AI type system to generate realistic drum tracks.

https://s.surveyplanet.com/mwahq3ut

Dave G.

Umbo Gumbo LLC/Music Power Studios

www.umbogumbo.com

[gentleclockdivider]

Yeah , no

[dgoldwag]

Why not?

[Ah_Dziz]

I'm not interested in the product or feeding data into your research on the product. Adding "AI" humanization to existing drum sequencers seems like it could be handy by doing things that real drummers commonly do. Like shifting the back beat at higher tempos and some more common rushing/ dragging behaviors. Having these things be "smart" when turning up a "humanize" knob rather than just randomizing things could be cool.

That being said, learning enough about drums to be able to program a decent drum track via samples and MIDI is a large part of composing music, and ignoring the things a real drummer does when programming drums on purpose is another large part of composing. I don't want an AI "writing" my drum parts. I don't really feel like helping out a company that thinks AI should be writing drum parts outside of maybe "auto accompaniment" scenarios. No offense intended.

[BBFG#]

Hmm, will it also simulate a drug addled depression spiral after getting dumped by its girlfriend?

[Ah_Dziz]

I'd buy that plug-in

[paramita123]

The website does not scream high-end AI capability to say the least.
It seems the people behind have real music credentials.

But from playing and producing to making yet another semi-AI-half-baked-JUCE-fueled drum program, and online on top of that, sorry but no. If reputable companies can't do it, I bet my money those gentlemen won't either.
But good luck anyway, if it works I'll probably test it.

[dgoldwag]

I appreciate everyone’s opinion on this most interesting topic.
Yes, we are musicians and a very small “possibly soon to be company”. We have many years experience writing, arranging, recording, producing, and mixing music from amateur to professional level. We also have a few accomplished drummers on our team.
Our motivation is to give everyone access to a cooler way to produce realistic drum tracks. We acknowledge that the “gold standard” is a real drummer in a high quality studio but for obvious reasons, this is not always available.
We have some concepts that the existing products don’t yet have (I’m being purposely vague) and we think doing it in the cloud will allow for seamless upgrading and improved capabilities of the instruments.
We are self financed (for now) and have to make decisions on whether this concept is viable and so
the survey was our first attempt to gauge the market.
Dave G.

[cake builder]

I think many people (including myself) are opposed to tools that take musicians out of the equation.

If it were a tool I could feed my MIDI tracks to augment and produce something more natural sounding, I'm not opposed to that. If it were something you give a prompt, e.g. 'give me a hard rock track at 120 bpm in 4/4' and it spits out a full drum track, I don't want anything to do with it.

[Funkybot's Evil Twin]

Here's my take on it: EZDrummer has "Bandmate" and "Song Creator" features already. They can listen to your audio, find a groove from a library that works through Bandmate. Then you can bring that groove into "Song Creator" and build up a song with it around pre-defined structures. Once you have that, you can give the virtual drummer instructions to change their lead hand from the hats to ride for example, or play more or less busy, etc. No one complains that this functionality exists. Writing drum parts can be tedious as hell, and that works to get users up and running quickly. The only difference I see is that it's based on a pre-existing groove library. It's also wildly hit or miss and comes up with some truly bizarre "bridge" grooves very consistently.

So yes, I'm 100% in favor of AI drummers. Here's how I think that could work: have real drummers on MIDI kits playing along to drummer-less music and creating drum parts. Do this in a bunch of different styles, tempos, etc., but have a focus on what the drummer does well. Feed that training data into an ML system for each drummer. So we can have a Classic Rock drummer, a Punk drummer, a Pop drummer, Modern Jazz drummer, bop and big band drummer, etc. Obviously, any drummer can play any style of music and that's where it could get interesting: throwing a bop and big band drummer in a rock song for instance. Basically: make Jamstix (another product no one complains about the existence of) and make an AI-based version of that.

Just pay the drummers that helped you build the training models (e.g. "session musicians" and make sure they're ok with what you're trying to do - maybe even give them a percentage of sales). Then you can expand the product by selling additional drummers.

I don't even need the product itself to have a drum library. I've got those. Just map to what's already out there: BFD, Super Drummer, EZ Drummer, Addictive Drums. Once the generated MIDI is in the DAW, I and other "real musicians/producers" can edit it as needed, but the goal is to get up and running quickly with realistic drums as a starting point.

I'm 100% in. Someone's going to be first to do this (seems like it might be Apple from the quick blurb I read on Logic 11 but that's locked to one DAW). The naysayers will say nay. And they can do things however they like. But again, Jamstix, EZDrummer, etc., pretty much every drum product offers tools to speed up creation of drum parts. Having an AI drummer isn't much different IMO. It's just a matter of "is this generated from a human via a groove library" or "did a human train an AI model to generate these grooves". The only thing that matters to me is how well it works and how much time it can save.

[Funkybot's Evil Twin]

Also, to the "learn how to program your own drums, it's better crowd", to which I include myself (this was my thinking for 15 years and exactly what I did), I'd also add this: that makes it really easy to under-estimate what drummers do out of ignorance about drumming. If you're not a great drummer, you're not going to write great drum parts. Like, do you understand the nuances of a six stroke roll and various rudiments along with when and why you'd use them? Probably not. But a good drummer would.

I can get as good results or better than EZDrummer with Song Creator but it just takes a hell of a lot longer for me to roll my own. And even when I use it, I always end up having to edit the performance in some way. But these are pretty basic drum parts. I'd love to get better results with an AI drummer. It's about the most benign usage of AI in music I can think of outside of amp software models.

[dgoldwag]

I think it's great that the survey stimulated some discussion about AI. There certainly is a potential threat to the "status quo" of making music by this type of technology but our concept is not particularly focused on the "it's a rock track at 140 bpm......go" capability although any drum track system should do that.
Our focus is to try to quantify and program those qualities of a real drummer that make drum tracks come to life.
I have built tracks using the standard virtual drum products by editing virtually every hit and the final product was good for a non-drummer but it took forever and was really tedious.
I have had many more tracks that I built with grooves and sounds from SD3 that came to life after my drummer came in and played the track. Sometimes, things he did on the track led me to go back and re-arrange other parts. He is a truly talented gospel/R&B drummer who rocks. My motivation for this product is that I can't have him in my studio every day but it would be cool if I had some of his mojo available.
I don't think there is much chance of us building a system that would totally replace a real drummer but I think there are lots of people out there like me who could use quick access to something like this.
Dave G.

[Ah_Dziz]

Also, to the "learn how to program your own drums, it's better crowd", to which I include myself (this was my thinking for 15 years and exactly what I did), I'd also add this: that makes it really easy to under-estimate what drummers do out of ignorance about drumming. If you're not a great drummer, you're not going to write great drum parts. Like, do you understand the nuances of a six stroke roll and various rudiments along with when and why you'd use them? Probably not. But a good drummer would.

I can get as good results or better than EZDrummer with Song Creator but it just takes a hell of a lot longer for me to roll my own. And even when I use it, I always end up having to edit the performance in some way. But these are pretty basic drum parts. I'd love to get better results with an AI drummer. It's about the most benign usage of AI in music I can think of outside of amp software models.

I'm a big fan of learning to make my own parts so that I can learn exactly why various rolls sound good where they do and the differences between all kinds of performance choices and how they interact with production choices. I'm happy with intelligent humanization of inputs and there are lots of cool ways to do this, including AI.
I'm generally not very interested in feeding market research for companies I don't know for products I know nothing about. Nothing personal to this company. Having useful tools for musicians is always cool. Hopefully these folks come up with such a tool.

[mixyguy2]

I'm also opposed to tools that take musicians out of the equation and I've never so quickly gotten so sick of an acronym (or concept) as "AI" in my life and nowhere more so than regarding music. AI musical tools are for people who lack the talent to do things themselves, and frankly I'd just as soon they admit that and not do it (or get someone else to do it). The road AI is about to take us down musically is one sad rabbit hole I don't even want to think about.

Also, it sounds like you're trying to do what Band in a Box is already doing.

[dgoldwag]

It's very likely that nothing I post will convince anyone who feels negatively about this idea so I won't attempt to change anyone's opinion, but I do want to emphasize that this concept is not designed as a replacement for a real musician playing a real instrument but rather as a very sophisticated tool for all musicians to use for creative purposes.

Not everyone has the budget or situation to track with a real drummer in a great studio so I think the best of new technology can bring capabilities that were limited by availability and budget to a wider audience.

Dave G.