How Photo Retouching Disappeared From My Studio in Five Years — and Branding Didn't

Five years ago, if you'd looked at our invoices at 35milimetre, most of them would have said "retouching" or "3D." Those two lines carried the studio. Today, almost none of them do.

That's not because we got worse at retouching, and clients didn't stop needing clean images. The work itself moved: most of what used to take a trained retoucher an afternoon now takes a client's own marketing assistant a few clicks. What surprised me isn't that retouching shrank. It's what grew in its place.

What retouching used to mean, and what killed it

Ten, even five years ago, cleaning up a product shot was slow, deliberate work: clipping paths by hand, dodging and burning to sell a highlight, cloning out a scuff on a bottle, matching color across forty SKUs shot under slightly different light. That skill was billable because it was scarce and it took real time.

Photoshop's 2026 tooling changed that math directly. The Remove Tool, rebuilt on a diffusion model instead of the old Content-Aware Fill, now handles the blemish, dust, and stray-object cleanup that used to define a junior retoucher's day. Object selection that took 30 to 45 minutes of careful brushwork now takes two or three, according to reporting on this year's release.

None of this means that retouching skill has disappeared from the world. It means it stopped being a standalone line item on most jobs. It got absorbed into the last five minutes of someone else's workflow instead of being invoiced as a separate service.

3D got smaller, not gone

We used to run a genuinely balanced shop: one graphic designer, one 3D artist, and me directing and compositing. 3D was close to half the business.

It's a much smaller market now, but it didn't vanish, and I don't think it will. What disappeared is the routine end of it: quick mockups for internal decks, placeholder renders for a pitch, product visualizations that just needed to look plausible for a moment. AI image tools do that well enough for those purposes, and fast.

What's left for CGI is the harder end: hero shots where a client needs pixel-accurate control over a material, a liquid, or a reflection, and where the render has to match reality closely enough to survive scrutiny. That distinction matches what I've seen written about the CGI-versus-AI split more broadly: AI output is probabilistic, useful for exploring ideas quickly, while CGI stays deterministic, which is exactly what you need for a final commercial asset. 3D didn't lose its value. It lost its volume.

The part I didn't see coming: branding and label design getting busier

Here's what actually grew at the studio over the same period: identity work, label design, packaging systems for FMCG and beverage clients. That's not what I expected five years ago.

I think the reason is fairly simple. A recent industry report found that 88% of businesses now use some form of AI design tool, yet 30% still describe AI-generated design work as lower quality than what a human produces, and brand identity and packaging remain the two project types companies most want a real designer handling.

That tracks with what I see on my own screen. AI can generate an illustration or a label layout quickly, and sometimes it looks fine at a glance. But in my opinion, most of it still reads as amateurish the moment you put it next to real shelf competition: the typography doesn't quite earn its confidence, the color logic doesn't hold across a product range, the whole thing looks assembled rather than designed. Even a designer with a couple of years of experience, working carefully, usually beats it on a full label or identity system right now.

Part of the reason is structural, not just aesthetic. AI tools are good at producing a design; they're not yet good at producing a brand, because they don't carry the cultural context, the retail logic, or the long-term coherence a brand needs across dozens of SKUs and years of updates. And when every brand in a category reaches for the same generator, the risk isn't bad design, it's identical design; nothing on the shelf stands out from anything else.

How this changed what we actually do

35milimetre is still three people: me, one graphic designer, one 3D artist, plus freelancers we bring in when a job needs more hands. What changed is the mix of work we take on.

Retouching-only jobs are rare now; when retouching shows up, it's usually the final polish inside a bigger compositing or CGI job, not something a client hires us for on its own. 3D is a smaller share of revenue, but the jobs that do come in tend to be higher-stakes and better paid, which is a fair trade. Over the last year we've also built AI into the front end of our own process: fast concepting, option-generation for clients who want to see direction before we commit real hours. The delivery layer for anything brand-facing, a label, a packaging system, an identity, still runs through the designer's hand and my direction, not a generator's output.

What this actually tells you

The tools didn't just make old work cheaper. They redrew which of our skills are scarce. Right now, brand judgment is scarcer than pixel-pushing, and I don't think that's a permanent law of the universe; it's just where things stand this year. For the time being, though, human-led branding and label work has more room to grow than it has in a long while.

If you're building a label or a brand identity system and want a second opinion from someone who's spent 26 years looking at what actually holds up on a shelf, we're happy to talk.