New AI retouching techniques in Photoshop: the 2026 workflow

Look, I rolled my eyes too when Photoshop started cramming AI into every panel. Most of it is marketing noise — but not all of it. After a full season shooting commercial work across Dehradun and Uttarakhand, I rebuilt my whole retouching pipeline around the tools that actually held up when a client deadline was breathing down my neck. Here's what survived, what didn't, and where I still reach for a brush.

The removal trio

First decision in any retouch session: which removal tool do you grab? Adobe ships three overlapping options now, and they're absolutely not interchangeable.

The Remove tool — the little band-aid icon — is my default for small distractions on clean backgrounds. Dust specks, sensor spots, stray wires, a random price tag peeking out of a seam. Brush over it once and it's gone. The fill is lightweight and works in real time, so you can scrub an entire frame in seconds. On e-commerce batches where dust is the main enemy, this thing is the workhorse.

Generative Fill is the heavyweight. Select a region and it regenerates the whole area from scratch using Firefly. Where it shines: removing a large object that's embedded in a textured background — a reflection in a mirrored surface, a pole running behind the product, a cable crossing the frame. The key is the selection. A tight, clean selection around the target gets you a vastly better result than a sloppy marquee, and I've learned that the hard way more times than I'd like to admit. Spend the extra twenty seconds on the selection and you save ten minutes of regeneration attempts.

Content-Aware Fill has quietly become the fallback, not the default. It still wins in exactly one narrow case: filling a large area that has to match the existing texture perfectly, like extending a marble surface or a seamless sweep background. Its pattern-based fill is more predictable than Firefly for pure texture extension, and it's faster than iterating on prompts. For everything else? Generative Fill beats it.

Selecting the subject

Masking used to be the most tedious part of product retouching. I'm not exaggerating when I say hours went into manually tracing bottle curves, glass edges, and reflective surfaces. The Object Selection tool changed that math completely.

For a product on a clean background, Object Selection with a single click now produces a mask good enough for ninety percent of compositing work. The trick is knowing the two refinement steps that follow: Select and Mask for edge cleanup, and the Refine Edge brush for hair-like or glass-like boundaries.

Glass and transparent products are the exception — and they'll humble you. Object Selection gets confused by glass because the AI has to decide whether the bottle, the liquid inside, or the reflection behind it is the subject. For beverage photography I still build glass masks manually. But I use Select Subject on the label area only, since labels are opaque and selectable, then extend the mask along the glass contour. It's faster than full manual masking and more accurate than trusting the AI alone.

Here's the trick worth learning this year: run Object Selection on the Background layer and invert. Select the background, invert, and you get a product mask that includes soft shadows automatically — because the shadow belongs to the background, not the product. That single technique wiped out a huge chunk of my manual shadow-retaining work.

Neural Filters: the useful shortlist

Adobe's Neural Filters panel is a graveyard of gimmicks hiding a handful of genuinely useful tools. After months of production use, here's my shortlist.

Smart Portrait is the most powerful and the most dangerous. The skin smoothing slider is excellent for lifestyle beauty shots where the client wants a clean finish — but it's dangerously easy to overdo. I cap it at twenty percent and always mask it so the eyes, lips, and hair stay critically sharp. The face-aware lighting adjustment is handy for evening out uneven light on faces in group shots.

Photo Restoration is genuinely excellent for bringing older campaign images back to life — removing scratches, JPEG artifacts, and film grain while upscaling. I've used it to rescue legacy product shots from previous brand campaigns that clients wanted repurposed for social.

Color Transfer is the one I use most in commercial work. Pick a reference image and it maps its color palette onto your shot. It's not a substitute for proper grading, but it's a phenomenal starting point for matching a new batch of images to an existing campaign look. Grade the anchor image by hand, then use Color Transfer to get the rest of the batch in the same ballpark before manual refinement.

The rest of the Neural Filters — JPEG artifact removal, Super Zoom, style transfer — are either covered by better tools elsewhere or too unpredictable for client work. I keep them installed. I rarely open the panel for them.

The batch pipeline

The real productivity gain isn't any single tool. It's how the tools chain together into a batch workflow. Here's the exact sequence I run on every e-commerce delivery.

First pass: cull in Lightroom, sync white balance and exposure against the color reference card, then export selects to Photoshop as layers or via Lightroom's Edit In workflow. This pass is unchanged from the pre-AI era.

Second pass: batch cleanup with the Remove tool. I record a Photoshop action that runs Remove on each layer, then process the batch through Image Processor or a droplet. Dust, spots, and small distractions disappear across every frame in one sitting.

Third pass: per-image creative retouching. This is where human judgment lives — deciding what to remove, what to keep, which reflection adds character, where the seam should sit. AI handles the mechanics. I handle the decisions.

Fourth pass: consistency review. I overlay all final images in a contact sheet and check for three things — color drift between frames, mask artifacts (halos, torn edges), and whether the retouching has erased any product detail that matters. Every frame gets reviewed by a human before delivery. The tools have never once flagged a subtle color cast that a client would notice. That job remains mine.

Where AI still fails

It would be dishonest to present this as a frictionless upgrade. There are three places the AI tools still cost me more time than they save.

Reflective and metallic surfaces remain the weakest area. Chrome, stainless steel, and liquid produce specular highlights that Generative Fill and the Remove tool simply guess at — and the guesses look wrong. A reflection that should be a clean, continuous streak comes back broken or warped. For anything with hard speculars, I remove distractions with the healing brush and rebuild reflections manually.

Text and logos are another failure zone. Remove a sticker from a bottle and Generative Fill often invents replacement text or warps the label geometry. The fix is the Clone Stamp tool combined with a tight selection — and patience. For label-heavy products, I plan the shoot to avoid covering label text in the first place, because fixing it in post is rarely clean.

The third failure is judgment, not pixels. AI can clean an image, but it can't tell you whether the cleaning serves the brand. A ceramic mug with a handmade glaze has texture that is a selling point. An AI first-pass will happily smooth it into generic perfection, destroying the artisanal story the client is paying to communicate. Every automated pass needs a human with the brief in hand, deciding what the product is supposed to look like.

The bottom line

The 2026 Photoshop retouching stack is genuinely transformative for one specific kind of work: high-volume e-commerce cleanup on matte, textured products with clean backgrounds. For that work, the Remove tool, Object Selection, and batch processing have cut my turnaround by roughly sixty percent.

For reflective products, glass, and brand-critical detail work, the AI is still a co-pilot at best — useful for the first pass, overruled on the final decisions. The photographers who get the most from these tools aren't the ones who trust them most. They're the ones who know exactly where each tool breaks, and that knowledge only comes from pushing them against real client work.

My advice to brands: ask your photographer which AI tools they use and where they still retouch by hand. The answer tells you whether you're getting a production pipeline or a magic-wand demo. The tools are common now — the judgment is what's still rare.