Toontracks AI development?

[Nsurround]

It is obvious to me that toontracks developers must be working overtime to train AI on its current ‘sampled’ products an dmore. IMO, unfortunately the music production world is moving heavily into AI tools and or ‘AI generative music’ and the competition is really heating up to develop these types of products.  I assume if Toontracks still wants to stay in the game that this is where they are headed. They already have produced some of the best sampled drum kits on the market along with many midi grooves. So it stands to reason that the most recent release of EZdrummer 3 and its ‘bandmate’ feature is just the starting point.  With this in mind, I wonder where this will end up taking its current products such as Superior Drummer 3 in terms of updates or new features. The other choice would be to add completely new product(s) that will try and more fully realize this AI revolution (Superior Drummer 4 . . . . . ).  Again IMO, there is a big difference using AI in the context of it being a ‘tool’ rather than a self ‘generative’ crutch (aka Suno and others). It remains to be seen how Toontracks will play this out. This road also encompasses Toontracks other non-drum music production products but used the drummer ones as the example. Any Thoughts?

1

Thanked by: ahirvo

Topic

[gearnerd]

The most likely area where they might be able to integrate AI is their midi tools. There is definitely room for improvement to paraphrase certain styles of playing to different chords and rhythm structures. It’s to rigid right now and breaks too easily. Also the controls in terms of the amount of notes and loudness work on a technical, rather than musical level.
I like Logic’s AI midi generation much better. Not necessarily the results but the concept and the controls it gives you using parameters like complexity, intensity and how it’s more resilient if you make structural changes.

For an AI future, I imagine, instead buying  midi packs you would buy a style or a “drummer”/“bassist”/“keyboard player” add-on.

[Nsurround]

“For an AI future, I imagine, instead buying  midi packs you would buy a style or a “drummer”/“bassist”/“keyboard player” add-on.”

Yeah, I think the ‘groove’ section could be done better and that the ‘style’ maybe the key. It seems to me that the ‘Bandmate’ feature in EZdrummer 3 is going in that direction. At least for songwriters who do not want to spend much time going thru a whole bunch of grooves via the current options. Seems to me that Superior Drummer 3 has the best sounding library of drum samples but it uses an interface of tweaking options that is somewhat cumbersome to mess with unless you are really into it. I think AI tools could improve that quite a bit and get you much faster to a satisfying end point. Since Toontracks has a infringement clause about using AI training on their products, I assume that the best defense is going on offense and creating the AI tools in their own eco system. Not sure that they can really stop others from AI training of their products unless they have some method of identification of samples and midi grooves etc. To me it would be extremely hard to do this at this point in AI development.

 

[BlueMistral]

How about using the GPU to process audio?

https://www.musicradar.com/music-tech/software-apps/what-the-gpu-offers-for-audio-is-an-almost-unbound-level-of-processing-we-investigate-the-claim-that-harnessing-gpu-can-unlock-limitless-music-production-potential

[ahirvo]

Editing drum and bass tracks is quite slow and tedious process, even with the help of Bandmate. The thing is that the generated grooves are rarely exactly what I want, and I need to use a lot of time for tweaking and adjusting the details (placing individual drum hits, finding right fills, etc.). It would be great to get next-level AI assistance for this. The technology is ready for sure, the solution just should be designed and implemented for Toontrack products.

Approach using AI chat prompts could work.

Example Use Cases (prompts):

1. Control Play Style

• Goal: Change style
• Context: Groove: hi-hat + snare pattern, Style: rock → funk
• Request: “Make the hi-hat more syncopated and add ghost notes on the snare.”

2. Edit Groove to Follow Guitar Rhythm

• Goal: Edit groove
• Context: Groove: Selected groove, Reference: guitar track rhythm
• Request: “Make the kick drum follow the palm-muted guitar chugs.”

3. Tempo Change with Triplets at Cursor

• Goal: Adjust tempo + timing
• Context: Cursor position: bar 16, Next groove: triplets, Reference: previous 4/4 straight section
• Request: “Shift to triplets so that the note frequency stays the same as the previous straight groove.”

[Nsurround]

Hey, I would agree that producing really good drum tracks based on  song writing or composition can be tedious even using the current ‘bandmate’ feature.  The AI prompt idea I think could work in the editing process once you have a basic groove going that fits with the song style etc. I like the idea as long as the human input is in control of the process. This is a great tool for songwriters and such but I would not like it to turn into something like Suno or other similar apps that seem to be somewhat currently popular. Those apps seem more of a novelty item, at the moment, than a serious production approach to music. That being said, it is all changing for recorded music and where it goes is not clear at this time. My hope is that  live shows/tours will keep steering the interest in music creativity and such so that recorded music is still driven for the most part by skilled or dedicated human musicians and producers. using in part AI as a ‘tool’ and not a ‘crutch’.

[Author]

Replies