Integrating Generative Design Tools into Tradition

[Highlevele]

Lately I’ve been experimenting with integrating generative tools (Stable Diffusion, Flux, etc.) into traditional design workflows, specifically for client deliverables that still require brand consistency and revision history.

The biggest challenge isn’t the output quality anymore — it’s managing
(1) control over micro-details,
(2) reproducibility across revisions, and
(3) how it hand-offs to execution tools like Illustrator, Figma, or print-ready pipelines.

Curious how others are handling:

• When generative assets enter the workflow (ideation vs. final assets)
• How brand constraints are enforced (brand kits, custom model finetuning, ControlNet, etc.)
• Revision control with clients who expect editable vector layers
• Whether these tools actually speed delivery or just shift effort
• Any industry-level pushback (legal, creative, or expectations)

On the Highlevele side, we’ve been testing structured pipelines to keep outputs consistent across campaigns — lots of promise, still lots of quirks. But it does feel like the gap between “experiment” and “production” is starting to close.

Would love to hear how designers here are approaching it and what’s working (or not working) in the real world.

[Smurf2]

Every major shift in design has gone through this same phase of discomfort and debate. From movable type replacing hand lettering, to lithography and offset print standardising reproduction, to desktop publishing disrupting traditional typesetting.

Each step was seen as “the end” of something, but ultimately just became another tool in the craft.

Illustrators and photographers felt it when clipart libraries appeared, then again when stock image CDs flooded studios, and later when online stock sites made high-quality imagery instantly accessible. None of those killed creativity they changed where value lived, concept, judgement, curation, and execution.

AI feels like the same pattern. It’s not replacing design thinking or brand stewardship, it’s compressing certain stages and shifting effort elsewhere. The challenge, as always, isn’t the tool it’s how well designers control it, integrate it into real workflows, and uphold standards clients already expect.

Seen through that lens, generative tools aren’t a break from design history they’re just the next step in a long evolution.

[mluxgd]

Hi There , Generative AI is not perfect, and it makes mistakes or makes defects. I have read someone says that “Generative AI” is like a slot machine, because you must with be luck in order to get the best results. For me, yes, I use, and I am still using Stable Diffusion, Firefly and other AI tools to generate content (images) but not anything. Sometimes you will need to make your own, since Generative AI can cause defects. These are examples of a Generate AI defects which I have to fix myself :

[Just-B]

Although it’s improving, AI is still not good enough, especially for deliverables.

It’s good enough to check for writing mistakes, but not good enough to produce writing that is natural and nuanced.

The imagery it produces can be amazing, but it usually looks fake and unnatural. Clients are increasingly able to spot it, and many don’t want it.

I’ve used it for background research, which is valuable, but needs to be taken with a grain of salt since it’s not necessarily reliable.

I’ve used it to both check and write JavaScript and Python scripts, which sometimes work and sometimes don’t.

I have no use for AI in ideation. I’d rather use my own knowledge and experience than go through a drawn-out iterative process of AI trial-and-error.

I’ve rarely designed under brand constraints. Instead, I’ve been more often someone who creates them, so I have no use for AI there, and I really don’t even see how it could be usable. I’m open to learning how it could be helpful, though.

I’ve never needed assistance with revision control. I’ve only allowed three sets of revisions, and when one is obsolete, it’s tossed into a folder that’s usually forgotten and never looked at again.

I would never trust today’s AI to give me what I needed, and I would spend more time checking for mistakes than if I hadn’t used it at all.

Photoshop’s AI is mostly worthless to me, except for generating generative fills.

Bottom line: it’s currently a nice tool to have in the toolbox for various things, but so far it’s hit or miss for this and that.

[PrintDriver]

Generative AI in Adobe products costs money now. Or rather, credits that you buy with money (after you run out of your allotment.)
Topaz is the same, if you want to use their render farm to get a result in something under a lifetime.

[Smurf2]

Never needed them before. Don’t need them now.