AI Related Suggestion

As an IT professional for many years, now a full-time composer and orchestrator and a loyal Cubase user, I have witnessed how disruptive technologies reshape entire industries. AI is one of those paradigm shifts.

I highly encourage Steinberg to make Cubase AI-native rather than treating AI as an external tool. Imagine importing a simple piano sketch (melody and harmony), then letting Cubase generate clean, editable MIDI tracks, suggest orchestrations, articulations, expression, and arrangement ideas; all while keeping every musical decision under the composer’s control.

The composer and orchestrator should always remain the creative force. AI should perform, assist, and inspire, not compose on the artist’s behalf. I believe this is the workflow that professional musicians are looking for, and Cubase has a unique opportunity to fill this gap.

In my opinion, professional composers will always want to create their own original music. They are looking for AI that enhances their creativity and productivity, not AI that replaces their creative role. Cubase has a unique opportunity to lead that future.

Mohamad Hejazi

Nightmare to pull off. Steinberg/Yamaha would have to lease incredibly expensive time from anthropic etc to dev the capability …hire astronomically-high salary coders who know how to work the underlying training…astronomical…stratospheric costs to pull & replace code every 6 months or so as underlying code advances change…severely jacking up the price of Cubendo into maybe thousands of dollars…or subscription territory…nughtmare to recoup investment…triple nightmare to ensure works flawlessly across multiple platforms such as with you know who in Cupertino.

And that’s the easy part for maybe one single solitary fairly straightahead instant ai “thing” feature that’s nowhere what your wish list is asking for for Christmas :slight_smile: imo imo imo

Now, you get out there to 2050 or so, yeah, the story is a bit diff. But back here in ancient 2026, there’s only so much you can expect imo.

I cannot imagine opening Cubase without knowing what I’m about to compose. Framework of the piece, instrumentation/arrangement, snippets of melody - all done before I even turn the PC on.

AI - the marmite of the tech world!

I’m the exact opposite, I have no idea what I’m going to create when I power up the computer…I grab the guitar while I’m waiting and an idea will hit me…or I may decide I’m doing a piano piece, etc. I load Nuendo and start fiddling around with the piano or guitar and just addon as I go…additive composing I guess :wink:

I can’t picture working any other way and I get pleasantly surprised sometimes…for me ideas are a dime a dozen, it’s finishing something that’s the hard part :slight_smile:

@Googly_Croft Marmite is indeed like AI - either you hate it or love it.

For me, too, and probably all of us.

@Reco29 , he gets it!
For the record, I don’t like Marmite, although my mother used to insist I loved when I was a toddler.

The problem with that suggestion is nobody knows where the data used to trained those algorithms comes from, nor if it was acquired legally with consent from the owners. Ask any AI company that and you’ll get a non-answer each time. When you attach a tool of questionable legality to a DAW, you’re opening a can of Pandora boxes. Music is already complicated enough without Active Incompetence “helping” making things easier.

Keep AI out of Cubase… If at all, provide these things as seperate installers, so that users can choose for themselfs

I can imagine being so lazy that I never have to get out of bed again but I wouldn’t describe it as a desirable way to live. Part of the enjoyment of making music for me is the striving, effort, and yes, sometimes downright frustration with the process of composition, tracking and mixing. Without that, where’s the fun?

Yup

Other than being internal to Cubase you can pretty much do all of this now (and free in my experience) using Claude. It can read in an audio or MIDI file do a harmonic analysis and spit out some ideas or MIDI files.

Claude and I discuss music theory, especially around modes and more exotic scales quite often. It’s like having you own private tutor. I also find it very useful to find Cubase menu items, preferences, etc. for stuff I know Cubase can do but don’t recall where to find it.

Well, to an extent, very true.

And from the looks of it, Keep Marmite out of Los Angeles! :slight_smile: Never heard of the stuff before. Looks like grape jelly infused with dirt…or motor oil.

That’s the same thing.

Because there is no such thing as enhancing your creativity. If that’s not a marketing grift then I don’t know what is.

fair.

Just speaking for myself, but there are like a jillion things out there that I know from experience do enhance my creativity.

We are humans; already being maximally creative. We can’t get any more creative. But we can sure be lazy and lusting after fame and fortune. Trying to skip over the work while grifting about taking the shortcut.

About discussing music theory with Claude: I am sceptical to say the least. What I know is that you can interpret a chord progression in many ways depending on the underlying theory and school. And it takes not only profound theoretical background knowledge to understand the train of reasoning behind every approach but also a good amount of embedded human intuition to get there.
However, I haven’t tried Claude, yet. But that’s an interesting thing to check out.

About trying Marmite: I once expected a delicious sensation on my tongue knowing about Marmite’s popularity in the UK. Instead, I was immediately drawn to the sink. Like I had just licked the sweaty feet of a big hairy fellow after a long shift in the fish factory. You got to like that. I have never tried again ever since.

Like AI - you either love it or hate it. There’s no middle ground. The other discussions here demonstrate that.

I’ll post a portion below - it was a fairly long and wide ranging chat. First let me set it up (this will get wonky). I fed Claude a screenshot of the Key Editor of some bars that I’d written. It was in C Phrygian but also had an extra F# added to the scale so there was a cluster of minor 2nds F-F#-G which added a bunch of ambiguity in the 4th/5th neighborhood. Initially there was some confusion on what was happening due to Cubase’s ‘habit’ of misspelling sharps & flats. This is from right after that got sorted out.

A few things worth noting about this specific 8-note scale:

It’s Phrygian with a chromatic 4th/#4 pocket The F–F#–G cluster is the only place in the scale with consecutive half-steps. Everywhere else the scale still moves in its normal Phrygian intervals (half-whole-whole-whole-half-whole-whole from C). So harmonically/melodically, you have one “hot spot” of chromatic tension around the 4th–5th, and the rest of the scale stays clean and modal. That matches what you found in the synth line — F# consistently resolving up to G — it’s functionally acting like a chromatic approach tone to scale degree 5, not a wholesale mode shift.

The bass’s F#–C tritone reading still holds Even with the enharmonic correction, F# and C are still a tritone apart, so that oscillation in the bass (bars 11–12) is real and still worth noting as a distinct device from the synth’s stepwise F#→G resolution.

Naming-wise This isn’t a “named” traditional mode (it’s not Phrygian dominant, not double harmonic, not a standard bebop scale) — it’s simplest just to describe it as Phrygian with a chromatic passing/neighbor tone between 4 and 5. That’s a very normal, controlled way to inject tension without abandoning the mode’s identity, which fits the “one hot note” character you’re actually hearing rather than the denser polymodal picture I described before.

A bit later

What the added F# does to this pattern

Without F#, the F–G dyad is a plain major 2nd — the “calmest” interval in the set. By inserting F#, you split that gap into two minor 2nds (F–F# and F#–G) instead of one major 2nd. So you’ve added a third “hot spot” to the mode, turning the one previously mild region into the most chromatically dense one. That’s consistent with what you’re hearing in the synth part — it’s not random alternation, it’s Phrygian’s natural m2/M2 shift plus an extra compressed cluster right at 4/#4/5.

The effect

Secundal dyads built this way give you a sound that’s dissonant but modally dissonant rather than chromatically dissonant — the ear tracks it as “colorful mode” rather than “atonal cluster,” because the pitches are still constrained to (mostly) your 8-note set. That’s a very Bartók/Messiaen-adjacent technique, and it also translates well to electronica — secundal clusters are common in ambient/IDM writing for exactly this reason: tension without needing functional resolution.

Is the synth voicing strictly adjacent scale-tone dyads throughout, or does it sometimes skip to a 3rd/4th and only use the 2nds as color at cadence points?

Much, much later (we’d also briefly discussed Monk earlier, so that’s not a random mention).

Bars 13–14: genuine stacked secundal harmony
This is the more striking passage. You’ve got Db3, Eb3, Ab3 (G#3), and Bb3 all sustained simultaneously for real duration — not passing tones, actual overlapping held notes. Stack those up: Db–Eb (M2) and Ab–Bb (M2), a perfect 4th apart from each other. That’s a quartal-secundal hybrid voicing — two major-2nd dyads stacked a 4th apart — which is a genuinely more sophisticated harmonic gesture than bars 10-11. This isn’t “chord + added 2nd,” it’s a real cluster-of-clusters, and it’s the moment in the part that would sound most identifiably “jazz harmony” rather than “modal melody with color tones.”

So you’ve actually got a progression in harmonic density across the phrase: looser, single-note-against-drone dissonance in 10-11, tightening into a genuinely stacked secundal chord by 13-14, then presumably loosening back out toward 15-16. That arc — tension accumulating into a real cluster rather than being uniform throughout — is probably a big part of why it reads as purposeful rather than random. Monk does this too: he doesn’t hold the same dissonance level the whole tune, he escalates into the “wrong note” moments and lets them stand out.

Impressive!
The first analysis matches what you would expect in a “standard” context whereas the latest analysis also takes typical artistic means into account that you would expect in specific Jazz contexts or personal styles.
The question is: “How much info was needed to point Claude into the right direction?” I don’t know how accurate the sample was in terms of microtiming, -tuning and dynamics. These are all indicators (among other aspects) that would automatically impact human interpretations of Jazz samples (pointers in the right direction).
It would also be interesting to see Claude’s analysis of a Franz Schubert song from Winter Journey or a harmonically complex Metal song as they all come with a different set of trigger signals.

Thanks for the example @raino!
I can see how AI can be helpful to analyze music by providing analytical options. The interpretation will still be left to humans and remain biased due to individual and unique backgrounds of the interpreters. That won’t change. But I can see how these “discussions” with Claude can be fun and fruitful :+1:

EDIT: Don’t give Claude the difficult stuff unless you want to see the lights flicker in your neighborhood. Or don’t tell them what you are up to… just kidding :wink: