Programmable based presets in xml/json/txt format

Cubase
1

I would love to see Cubase 16 introduce a supported way to programmatically create and modify presets, FX chains, and their settings.

The interesting part for me isn’t simply adding “AI” to Cubase. I’d rather see Steinberg expose a structured automation/API layer that external tools, scripts, and LLMs could work with.

For example, today I can search Google or YouTube for a vocal chain and manually recreate it.

But an LLM can already take my notes, something like:

“Pro-Q 4: high-pass at 80 Hz, cut 300 Hz by 2 dB.
1176-style compressor: 4:1, medium attack, fast release, around 4 dB gain reduction.
De-esser around 7 kHz.
Plate reverb on a send…”

…and turn that into structured data.

Imagine if Cubase could then accept something like that and actually:

  • Create the FX chain
  • Insert the correct installed plug-ins or suggest the stock equivalent
  • Set their exposed parameters
  • Create FX sends and routing
  • Save the result as a preset or Track Preset
  • Recall and modify it later

Even better, this shouldn’t have to be LLM-specific. Give us a documented scripting/API layer and let developers build whatever they want on top of it.

Cubase already knows an enormous amount about the installed VST ecosystem. The missing piece is a safe, supported bridge between structured instructions and the DAW itself.

For people with large plug-in collections, this could turn “I know what chain I want” into “build it for me,” while still leaving the creative decisions with the engineer.

That would be a serious workflow time-saver and a AMAZING launch pad for “starting point”

We could Call it “DreddieChain” :sweat_smile:

2

I think I could quite possibly create this chain manually faster than it would take to write the instruction.

3

I agree… but at the same time, I wanted to create a tool I can help others who have ADHD like myself

Those who like to chase squirrels in the middle of something you were excited about…

Maybe I will write something open-source to do what we are talking about…. Based on my experience and my scratchpad of pausing YouTube interviews :relieved_face:

But it would be nice to take what you said and have CuBae-GPT say: “Got it fam… Go chase that squirrel, and I will ping you in 2 minutes

4

It’s a good idea but not for me, I’d rather spend that time creating music :slight_smile:

5

Yes sir that’s what I am trying to get back to… managing the randomness of the mind… being overly critical of settings/config… overcoming “analysis paralysis”

6

That could be a thing to be done after the MIDI Remote API is more useful for the end users instead of being a thing only integrators or people willing to spend time programming every aspect of the controller integration, including the interface representation.

7

This is where AI comes in. I’ve seen videos where you can connect Ableton’s API/MCP calls to an LLM.

So technically, you can take a picture of your notes on a napkin or scratchpad, and it will populate those values while you pee.

Honestly, I have my workflow, and I know where everything is…

But… my wife is a singer, and she hates this part of the process. She has an idea, but I am at work. Now she’s stuck.

She’s already frustrated with me cause i’m frustrated with her not following directions, and she can’t hold the FaceTime camera still :sweat_smile:

That’s just one funny use case :upside_down_face:

8

I’m gonna present your post in this order and say:

And who do you think is still gonna get the short end of the stick when the AI doesn’t work as expected? You. And who’s gonna have to fix the AI when it stops working? You.

You’ve just added another step of friction to your and your wife’s workflow. Because there’s a whole lot of talking about all the incredible aspects of LLM and the magic it can do and essentially none on when it doesn’t work, which is not a rare occurrence, especially when said tool is in the hands of non-tech savy people who don’t have a deep understanding on how to make it work.