I had a shoot come back from a client last month with a note I’ve seen a hundred times: “Can you just clean it up a little?” The background had a half-visible light stand, the framing was too tight on the left side, and the sky looked flat. Any one of those fixes is easy. All three, on a deadline, across a full gallery? That’s where you start feeling the friction.

I went looking for a faster path and landed on this one-minute breakdown from Aaron Nace at PHLEARN. What caught my attention wasn’t any single tool – it was the sequence. He chains four techniques together in a way that respects how these problems actually layer on top of each other in real portrait work.

Removing Background Clutter Without Painting Yourself Into a Corner

The first problem Aaron tackles is background distractions, and he goes straight to Generative Fill. The key move here is selection-first thinking. Before you fill anything, use the Lasso tool to draw a loose selection around the distracting element – give yourself a generous buffer of surrounding pixels, not a tight trace of the object. That buffer is what gives the AI enough contextual information to generate a plausible replacement.

Once selected, open the Contextual Task Bar (it floats below your selection automatically in recent Photoshop versions) and click Generative Fill. Leave the text prompt empty. That’s not laziness – an empty prompt tells Photoshop to match the existing background rather than invent something new. For plain backgrounds, studio walls, or simple environmental backdrops, this works surprisingly well on the first or second generation. Click through the three variations Photoshop offers and pick the cleanest result.

The output lands on its own generative layer, which means you’re non-destructive from the start. If a client comes back two weeks later and wants a different treatment, you haven’t flattened yourself into a corner.

Fixing Tight Framing With Generative Expand

Cropping mistakes happen. Sometimes you’re shooting fast, sometimes the client changes the aspect ratio requirement after the fact. Either way, you end up with a frame that’s too tight and no extra pixels to work with.

Aaron’s fix here is Generative Expand, which operates similarly to Generative Fill but works outward from the canvas edge. Use the Crop tool and drag the canvas boundary beyond the existing image – don’t check “Delete Cropped Pixels.” Then, with that extended canvas selected, run Generative Fill on the empty area. Again, leave the prompt blank and let Photoshop read the surrounding scene.

The results are better on images with clean edges (sky, simple walls, grass) and less reliable on complex architectural detail near the frame border. For portraits shot against environmental backgrounds, it handles the expansion well enough that I’ve stopped agonizing over tight crops during the shoot itself, which is worth something.

Sky Replacement as a Finishing Move, Not a Gimmick

Sky Replacement sits under Edit in the menu bar and it still gets dismissed as a novelty tool. That’s a mistake. Aaron positions it correctly – not as a first-pass fix but as a polish step after the background distractions are already gone.

Go to Edit > Sky Replacement, and Photoshop will auto-detect the sky region and mask it. The built-in sky library gives you a range of options organized by type. The slider that most people overlook is Foreground Lighting, which adjusts how much the replacement sky’s light characteristics are applied back onto the subject and foreground. Push it too high and your portrait subject suddenly looks like they’re lit by a completely different sun than the one in the original frame. Aaron keeps it subtle, and that restraint is the difference between a plausible result and a composited look.

The tool outputs a Sky Replacement group in your layers panel with the mask and color adjustment layers separated, so you can refine the mask manually if the automatic edge detection misses hair or complex foliage.

Camera Raw as a Final-Pass Blending Layer

The last step is one I’ve been doing for years but don’t see discussed enough: opening the flattened composite into Camera Raw as a finishing filter to unify everything. In Photoshop, go to Filter > Camera Raw Filter. This is where Aaron pulls the overall edit together – color grading, exposure balance, and a touch of tone work that makes the generative fills, the expanded canvas, and the replaced sky feel like they came from the same image.

The specific controls worth focusing on here are Temp and Tint for color consistency, the Tone Curve for contrast shaping, and the HSL panel if the sky replacement introduced any hue conflicts with the subject’s skin tone. A small amount of texture reduction in the Detail panel can also help the generative fill regions blend more smoothly with the original pixels.

Where This Workflow Has Limits

I want to be honest about one scenario where this chain breaks down: high-end commercial work where the client or art director is going to pixel-peep the output. Generative Fill is probabilistic. It produces plausible results, not accurate ones. On a low-traffic social post or a web-sized image, that’s fine. On a full-page print ad where someone is going to zoom into 300 percent, I still do manual frequency separation cleanup on any area near the subject.

The other friction point is consistency across a multi-image sequence. Each Generative Fill call produces a unique result, so if you’re running this on 40 frames from the same shoot, the background fills will have slight variation between them. For editorial work that’s usually acceptable. For e-commerce where the brand needs visual uniformity, you’ll want to standardize one fill result and use it as a template layer.

The core insight from this tutorial is that sequence matters as much as the individual tools – remove distractions first, fix framing second, replace the sky third, then unify with Camera Raw last. Doing it out of order creates compounding problems that are harder to fix.

Watch Aaron Nace’s full video for the visual demonstration of each step, especially the Generative Fill selections and the Sky Replacement masking, which are much clearer to understand on screen than in text.