The Future of Automation, AI and Workflow in Cubase - Where are we?

Cubase
, ,
1

I’ve been using Cubase for a long time and I’m starting to seriously wonder where Steinberg is going with automation, scripting and AI-assisted workflows.

Don’t get me wrong, I love Cubase, but I’m turning 42 soon and I work professionally with production, recording, vocal editing, mixing and mastering. One thing has started to become very obvious to me. I don’t want to spend the next 20 years doing thousands of tiny mouse movements for things that really don’t require any creativity.

I’m not talking about AI writing songs for me or producing music for me. I’m talking about all the boring repetitive stuff around the creative work. Routing 50 tracks, copying the same EQ settings to 170 channels, setting up groups and sends, renaming and organising tracks, repetitive timing edits, or going through a vocal and fixing obvious pitch problems one by one.

After enough years you actually start feeling all that tiny precision work physically too. Clicking, dragging, zooming in, zooming out and doing the same thing again and again. That’s where I think DAWs are going to have to change quite dramatically.

Why shouldn’t I be able to tell Cubase:

“Take these vocals, create the groups, route everything correctly and load my normal vocal setup.”

Or: “Analyse this lead vocal. Leave the notes that are already good alone, keep the natural vibrato and slides, but fix the obvious pitch problems.”

Or simply: “Apply this EQ to these 170 tracks.”

And what makes me question this more now is that parts of this are already happening elsewhere.

REAPER for example has ReaScript and a pretty extensive API. Because of that, independent developers are already building MCP integrations where an AI agent can actually control REAPER, create tracks, handle routing, insert FX, set up sidechains, work with MIDI, mixing and so on. That’s not even Cockos building the AI themselves. They simply made REAPER open enough for other people to build it.

Ableton has had the Live API through Max for Live for years, where developers can access and control tracks, clips, devices, mixer parameters and large parts of a Live Set.

Logic is already using machine learning for things like its pitch correction.

So for me the interesting question isn’t really whether AI will become part of DAWs. It clearly will. The bigger question is how open the DAW is going to be when it happens.

Cubase has macros, Project Logical Editor, MIDI Remote and a lot of clever functions. But compared to something like REAPER, it still feels very closed when you want to build something that can actually understand the project and perform larger workflows for you.

That’s the part that worries me.

Because I really like Cubase. I know it extremely well and I’ve spent years building my workflow around it. I don’t want to change DAW.

But I also don’t want to stay in Cubase just because I’ve always used Cubase if another DAW suddenly lets me remove 30, 40 or 50% of the repetitive work from my day.

For me this isn’t really about AI hype. It’s about workflow, and to some extent also about being able to keep doing this for a long time.

I want to spend my time listening, producing, making decisions and being creative. Not moving a mouse three millimetres back and forth thousands of times every day.

So I’m genuinely curious what other long-time Cubase users think. Has anyone here started looking seriously at other DAWs because of scripting, automation or AI integration?

And maybe more importantly, has Steinberg said anything concrete about opening Cubase up more in this direction?

Because for me, this is probably becoming one of the biggest questions around whether Cubase will still be my main DAW five or ten years from now.

2

Your post is way to long to do a full response but routing a bunch of vocal tracks into a group takes me seconds, and it is even faster since v15 came along. Applying 170 eq settings to these 170 tracks is Q-link. I find your post strange, if you are looking for something “better” just go explore other software. Cubase to me is pretty far ahead, many DAW´s still do not even have ARA support.

And Steinberg won’t announce future features.

3

Maybe you didn’t read it after all. Did you get my point?

Ara support is great. That’s not what I’m writing. There’s alot of manual work that I’d like to know if stenberg will let us automate.

5

No.

Not me.

You don’t have to use the mouse so much if you get some control surfaces. AI is not the answer to that problem. Also, you might be exaggerating just a bit.:man_shrugging:

You don’t need AI to do any of that. It can be done with a template.

That’s takes you out of the creative loop–and it’s kind of lazy.

You can select 170 tracks and apply the same EQ to them without AI.

There are already posts on the forum with people asking for AI that will make creative decisions and write tracks for them. If it’s happening elsewhere and you need it–you could go elsewhere and use it.

It’s not an interesting question. A digital audio workstation isn’t needed to generate synthetic music.

If your workflow is really 30% - 50% repetitive work–you may just need to revise your workflow.

6

what is AI workflow in a DAW? why do you need it so much? why are you applying same eq to 170 tracks? there is no standard correct way to route or do aux sends. it really all depends on your own way of doing it and what you are trying to do. A lot of DAW work is manual and thats what makes it fun and useful. I also use FL Studio and it has a AI help chat bot that is useful. It’s handy and faster than sorting through the detailed manual. I think thats more than enough use for it in a DAW.

If my luddite understanding of what AI and LLM is, it is something that needs to be always on and constnantly trained and updated all the time – if so, then having it integrated into a DAW product would mean it is constantly being trained on what you are doing all the time with the product, no? why would I want that?

7

Cubase 15 is a beast. We can all do pretty well anything in it that’s required for today’s music scene. Anything. For any genre.

If you’ve spent some serious time in other DAW’s, coming back to Cubase makes you realize that none of us even utilize half of the possible offerings built into it. You can literally discover something amazing about it every day that will save you countless hours of tedious work.

Like many others, I wouldn’t be quick to make a move to any other DAW.

8

I think there is some use for AI in a DAW, maybe not to use every day, but it can certainly be helpful in getting the gruntwork done.

Of course whether it’s useful or not, or will save much time, depends on an individual users’ workflow, and the size/complexity of their projects.

I work in orchestral TV/Film music, and AI certainly has potential benefits in that respect. VSL have already introduced something in their latest version of Vienna Ensemble Pro that demonstrates this.

With a single prompt, you can ask the AI to create a project with the instruments you want, all coloured, routed to buses/folders and named according to the instrument loaded. What would normally take easily and hour or two, takes a minute of watching AI do if for you, so you can get straight to writing music.

Inside a DAW, something as basic as ‘name all tracks and events according to the VST instrument preset loaded, and colour by instrument type’ would be very useful.

For in smaller projects, or for those who can just use the same template for most projects, it’s probably not as useful.

9

a lof you what your describe is template work. if there is a particular way you like to color things or set them up, you can use a template to remove the boring parts.

10

I’m turning 66 soon and I work professionally with production, recording, vocal editing, mixing and mastering, and I don’t want to learn a load of AI nonsense. There’a a lot of sense in doing things manually sometimes.

:slight_smile:

11

I’m going to go against the grain here and agree with the tenor of the original post:

I’m an experience Cubase used (have used it since it was 24/III on the Atari ST), and use Cubase professionally for hours daily. Yes, some of the things the original poster mentions can be solved with templates and workflow refinements.

But I think their larger point stands - I too spend the majority of my time on tedium to make the machine make the kind of music I want. I would go as far as to say that at best, somewhere around 20% of my time is spent on genuinely novel creative work, and the rest on clicking around to make the system go.

Why shouldn’t I be able to tell it things like: Take these raw vocal recordings, clean them up, take a first pass at comping them to assemble the most emotional vocal delivery, and create a vocal chain to make them sound like the lead vocal on [insert your favorite track by your favorite artist].

It would save me hours of work, and takes nothing away from my human creativity - it amplifies it since I could then use those hours and take whatever the machine cooks up and refine it to make it work for whatever I’m working on.

And I also agree that having an open API for your DAW will leverage the ecosystem of other technology out there to make it do things it can’t. Why not leverage the creativity of technologists who will build things you don’t have the resources/time/etc for?

12

I appreciate all of your replies, and reading the replies, I think my original post probably focused too much on examples that can already be solved with templates, macros or a better workflow. So I understand some of the criticism.

I absolutely agree with the point about doing things manually. I don’t want AI involved in everything. A lot of the manual work is part of actually producing music, and I enjoy that.
What I’m talking about is the work where doing it manually adds very little.

Vocal cleaning and editing is a good example for me like Timo wrote.
I spend a lot of time in recording sessions and after with plugins such as variaudio and melodyne doing tiny pitch, timing, transition and note separation edits. Of course the singer should give you the best performance possible, and I still want to decide how the vocal should sound. But after many years, thousands of those tiny movements genuinely take a toll on my arm. No template really solves that.

I do most of the chain myself. I sing, write songs, music, produce it, recording vocals, editing, tuning, mixing and mastering, including work for artists, film, TV and games. So all those small repetitive jobs add up since there’s a ton of manual workflow there we cannot template away.

The VSL example above is actually very close to what I mean. If something that normally takes one or two hours of setup can be described in one prompt and done in a minute, why wouldn’t I want that option?

And this is also where I think AI differs from a template or macro. Those are great when you already know what you’re going to need. I’m more interested in situations that come up halfway through a project, where the software can understand the current session and help with something I never prepared a shortcut for.

The same could apply creatively without giving up control. Cubase already knows things like chords, voicings and has VST like omnivocal and others. I’d love to say, give me three backing vocal harmony options for this chorus. Avoid doubling the lead too much, keep them in a realistic range and make the top harmony simpler.

Then I choose, edit or reject them. Or on the technical side, check these 40 backing vocals for missing sends, inconsistent routing, clipping or anything else that differs from the group.

That doesn’t replace me as the producer or engineer. I still make the decisions. I just don’t necessarily need to perform every small mechanical operation after I’ve already decided what I want. I think that’s the part I didn’t explain well enough originally.

13

I think clickbait titles are problematic.

Titles that intimate an adversarial or competitive target don’t actually contain the info needed to express what the user is looking for.

Express your title (edit this one maybe) in a neutral way that doesn’t excite emotion, and your intent will be better understood and less verbose – resulting in easier engagement by people who are interested in improving the program.

14

The OP title raised a question comparing Cubase to other DAWs, and then the poster gave specific examples of what those other DAWs can do that Cubase can’t.

Agree or disagree with the OP content, I don’t see anything wrong with the title - it raised an interesting question, IMO, and accurately suggested what the rest of the OP would discuss.

With all due respect, “Clickbait” is, of course, a derogatory term, and, personally, I don’t think the title deserves that.

15

First off I’m going to reply in good faith behind your intentions and some decent points you make, but I use LLMs enough for work (because I have to) to recognize your post was clearly written by or assisted by AI - fine, if you find writing to be boring and repetitive. But it can be hard to take such posts seriously or authentically on a user forum.

That said, I agree with your general point about spending more time on creativity and less on technical tasks. All along since the beginning of AI I’d been saying that’s what these tools should have aimed to do in the first place rather than attempt to replace human creativity, so it’s good that we are seeing tools move in this direction.

I also agree with opening up the API for scripting, but you seem to be conflating that with AI and this example which is an LLM chat:

This isn’t remotely the same as:

None of these are examples of your imagined LLM scenario, where you tell a chatbot what you want and it just magically does it all for you. They’re completely different technologies and back-end development tasks.

So let’s entertain this LLM chat scenario built into the app, since I feel like that might be your main request. Therefore, if we put the API scripting stuff to the side (again, different tech, different implementation), that leaves us with AI integration as you imagine it:

So I have to ask, what DAW is currently doing this? I’m not a developer but I don’t think it’s as easy as just slapping a chatbot into the app, so I don’t think the current DAW market is promising such a thing.

Now let’s talk about LLMs and their inclination to hallucinate or screw things up. As mentioned, I have to use some of these tools for my day job, and even with the most up-to-date tools it really is a crapshoot. I might ask gemini to format a document and add bullet points or bold headers and sometimes it works great, but the other half the time it deletes content, it INVENTS or CHANGES content, and any other number of odd mistakes. So I try to adjust my prompt and correct it, which sometimes only makes it worse. An hour later wrestling with a chatbot, I realize I could’ve just done the formatting myself.

I also have to use Adobe Premiere (video editing software) for work, and since they do have an open API I tested a chatbot integration for some very basic tasks. Same in video, there’s a lot of repetitive technical tasks that aren’t creative which I would love to save time on. But it was unbelievable what a mess it made of my timelines (thankfully I had my project backed up). It not only failed to execute VERY simple tasks, it actually made things worse.

This isn’t to say that tools can’t improve, because obviously they can and will. But they make a lot of mistakes and as for me, if I need to route 170 tracks or make precise pitch adjustments or do a bunch of automation, I don’t trust AI to do it correctly.

Also, speaking of automation, if you’re referring literally to audio automation, I consider that to be a creative task most of the time. Is it tedious? Yes. But there’s an art to it. No two people will automate something the same way, and there’s no hard and perfect science to how something like volume or FX or sends ought to be automated. It’s all about ears and taste, and that’s something I don’t want to hand over to AI.

For the record I’m not against your premise of a tool that assists with repetitive technical tasks so you can focus on the creative: I’d love to see this too. But I don’t think slapping a chatbot into the app is the solution, nor do I personally believe any of the major DAW devs are planning to do this within the near future.

In fact it would be more likely to develop an all new DAW from the ground up with this kind of implementation, rather than try to wedge it into 30+ year-old codebases. For example Suno Studio etc.

16

Hi @albinloan,

if you don’t want this, then just don’t do it in this kind of extremely exhausting way - if that’s economically possible. You could use Logical Editor even more often, also different - or at least more efficient - workflows, more efficient (reductionist) templates and even partitioned templates, more efficient project versioning, etc. All of which also takes valuable extra time, but might still be worth it. Working a DAW on a daily basis can very well turn into “slavery”, and no DAW can help you avoid all of it.

Sadly, more AI controlled macros and AI controlled automation lanes wouldn’t make anything easier for you, as far as creativity is concerned. Why? Complex DAWs like Cubase simply cannot become “easier to handle” with even MORE complexity - which is exactly what more AI controlled macros and automation routines would introduce. The hard truth is: you may need to find better ways to reduce complexity also around your daily editing work, perhaps better optimize your (certainly proven) workflows, and perhaps also revise / rethink your current ways of building up your studio projects.

And it’s true: of course PC workspaces themselves are pretty demanding, all currently available methods of interfacing are putting quite some strain on the human body - I get it. But then there are ways to at least partially compensate for that, by using more (or better) haptic controllers, by taking more rests, by doing some gymnastics in between, by considering to change diet, by adapting to different demands of your body, as you age. The human body was simply not made to do all these things, while sitting on a desk or console.

(annotation: this is absolutely not intended as an arrogant form of “advice”, hope you won’t mind my direct tone)

Cheers,
Markus

17

Yeah, fair points, especially about reliability. I’d agree that I wouldn’t trust a chatbot today to blindly make hundreds of changes without some kind of validation, preview and undo system. But I think there is one part where we’re talking past each other a little.

I don’t really see scripting / API access and an LLM as two unrelated ideas. The API is what gives the AI safe, defined things it is allowed to do, so instead of an AI ‘magically’ perating Cubase, it could create track, create send, read routing, insert plugin, set parameter, analyse selected audio and more. Then I describe what I want in normal language and the model translates that into those defined operations.

The chatbot is actually and almost the least interesting part. What matters to me is whether Cubase eventually exposes enough of the project in a controlled way that tools like this can be built around it, but your Premiere example is actually a good argument for doing it properly. I wouldn’t want an agent making unchecked changes either. I’d want things like preview, validation, limited permissions and easy undo, and I also agree with you about automation in the musical sense. Volume rides, FX moves and things like that can absolutely be creative decisions. I wouldn’t want AI deciding all of that for me. What I’m mainly interested in is the stuff around those decisions.

There are already third-party tools for Ableton (Live API/Max for Live) that can control large parts of a session, create tracks, write MIDI, arrange sections and basically build out a whole song structure with our samples we’ve created from a prompt.
I’m not saying we should create music with AI, but you get the idea, and Ableton didn’t have to build all of that. The important part is that the software is open enough for other people to build it on top.

Serum is another interesting example. Steve Duda has talked about opening Serum 2’s preset format so 3rd party tools could build on top of it, including things like AI-generated presets. Serum has also already been licensed in a headless form for AI-based cloud music generation, which ofc also is just the beginning.

The interesting part isn’t necessarily Xfer building an AI assistant themselves. It’s making the software open enough for others to build intelligent tools around it.

Today FL Studio already has Gopher doing actual project actions from natural language amd that’s just the beginning.

Reaper is open enough that third-party MCP tools can create tracks, route signals, insert plugins, edit MIDI, write automation and render.

Pro Tools has an official scripting SDK, and there are already MCP projects being built on top of it.

Vienna Ensemble Pro has an official MCP server for AI assistants.

Logic is taking a different route tho, but is integrating more AI directly into production tasks.

So of course, we in Cubase can do most of the individual tasks manually. The difference is that today it still doesn’t expose enough control for an AI agent to understand a request and reliably carry out those operations across the project.

That’s the gap I’m talking about and the gap is real unfortunately.

EDIT:
Yes, I do use AI quite often to help me write, mainly for two reasons. Sometimes I’m walking and just dictate my thoughts, and sometimes my right arm hurts, so typing a lot isn’t ideal.

I don’t really see anything wrong with that. The thoughts and opinions are still mine, AI just helps me turn them into readable text faster. To me, that’s kind of the point. :slight_smile:

18

The intention wasn’t to make it adversarial, but simply to compare where Cubase is today with what is already happening elsewhere. I can of course adjust the title if you think another wording would better reflect the discussion. What would you suggest?

19

Thanks Markus, and I don’t mind the direct tone at all. I agree with quite a lot of what you’re saying.

Where I think we see it differently is the idea that AI necessarily adds more complexity. For me, the interesting part is almost the opposite. Cubase is already incredibly complex underneath. I don’t necessarily want another layer of menus, macros and systems that I have to learn and maintain myself.

A good example is a vocal recording session. Some of mine can end up close to 200 tracks or takes, sometimes more.

Before I can even properly mix them, there can be hours of fairly mechanical work like organising takes into verse/pre/chorus, trimming away dead space, talking, paper noise and other junk between performances, cleaning clicks and noise, comping, tuning, timing and aligning doubles and harmonies.

I think every recording producer can relate to this, but of course, Istill want to decide which take has the right emotion, how tight the vocals should be and how much tuning is appropriate. That’s the creative part and fun part, but imagine being able to tell Cubase something like: clean this recording session up, organise the takes by section, remove everything that clearly isn’t a performance, flag anything noisy or clipped and prepare it for comping.

Then after we choose the takes, tune these naturally, keep the vibrato and slides, and align the doubles to the lead without making them sound robotic.

So yes, I can keep improving my workflow, and I should. But I also think the software itself can evolve so that fewer repetitive tasks require me to build the exact macro or template beforehand.

Dont you agree? :slight_smile:

20

You’re under no obligation to change it, but I would simply write something like Modern Automation and Workflow Improvements

22

I’ll meet you half way. :slight_smile: