You shot your product photos on your phone or basic camera setup. The lighting is uneven. The background is cluttered. The shadows look wrong. You can't afford a studio or a retoucher, but you need these images to sell.
Generative Fill in Photoshop can fix this in minutes per photo. It's not magic—it's a practical tool that removes distractions, fixes backgrounds, adds missing shadows, and creates consistency across your catalog. The catch: you need the right prompts and the right process.
This guide shows you exactly how to use Generative Fill to make amateur photos look intentional, polished, and ready for e-commerce—without learning Photoshop deeply.
Pay once. Keep forever.
Turn raw product photos into studio-ready catalog images at 15–20 per hour using Photoshop's Generative Fill — with prompts actually written for how Adobe Firefly parses text, not generic AI advice. This guide delivers 12 copy-paste prompt
What's included
Or get free updates & new releases:
Follow for updatesOlder background-removal tools leave halos. Traditional cloning looks patchy when you try to extend shadows or fix perspective. Retouchers cost $5–15 per image. Generative Fill understands context. You select what you want changed, describe the fix in plain language, and it generates a natural replacement that matches lighting, reflections, and perspective. For product photos specifically—where backgrounds need to disappear and surfaces need to look pristine—it's faster than any other single-person workflow. The real speed comes from using the right prompts. A vague prompt like "clean background" fails. A structured prompt like "seamless white studio backdrop with soft drop shadow, consistent with overhead lighting" works. You'll learn the difference.
Fast-Track (5 min/photo): You need consistency fast. Use a single standard prompt for all images—same background type, same lighting direction. Best for cataloging 10+ items shot in similar conditions. Precision (8 min/photo): Each product has unique challenges. You adjust the prompt for reflective surfaces, fabric texture, or shadow direction. Best for launch photos where quality matters more than batch speed. Batch Consistency (12+ images/session): You process multiple images with tracked adjustments so they all match even though they were shot separately. You'll build a reference image, lock your prompt parameters, and apply the same approach across variations. Best for scaling a product catalog over weeks. All three use the same four-part prompt structure. The difference is iteration and refinement between images.
Jewelry with gold and glass: Initial prompt tried "shiny background." It generated a reflective surface that didn't match the product's own reflections. Second prompt added specific language: "matte white, no reflection, with subtle shadow showing vertical light source to match product highlights." Result: clean, professional. Flat-lay with multiple items: Crowded spacing made backgrounds look busy. The fix wasn't just removing background—it was expanding white space while keeping consistent shadow depth across all products. Phone-camera apparel shot: Wrinkles and dull lighting. Generative Fill added soft depth shadow and corrected the fabric tone to match the actual color under proper lighting. Took 6 minutes. Each case study includes the failed prompt, the iteration notes, and the exact final version that worked. This saves you the trial-and-error.
Generative Fill uses Firefly, Adobe's image generation engine. It responds to a specific grammar: [Surface/Background Type] + [Lighting Direction & Quality] + [Specific Defects to Remove] + [Mood/Style] Example: "seamless white studio background with soft diffused lighting from above, remove shadow under product, add subtle drop shadow 2 inches below item, modern e-commerce style" Word count matters. 15–25 words is the sweet spot. Too short and it guesses wrong. Too long and it splits focus. Punctuation: Use commas to separate instructions, not periods. Periods signal a new concept and Firefly treats them as separate requests. You get a cheat sheet with 12 ready-to-use templates for jewelry, apparel, glass, shiny metals, ceramics, and flat-lays. Each includes the exact selection method (how to choose what area to fill) so you're not guessing about which part of the photo to change.
Distorted reflections: The generated background doesn't match the product's own shine. Fix: Lock the product in the selection, expand the selection outward only, and specify "no reflection in background" explicitly. Unnatural edge blending: The filled area looks pasted on. Fix: Feather your selection by 5–8 pixels before filling, and add "seamless blend" to your prompt. Wrong lighting direction: The shadow points the wrong way. Fix: Look at the product's own highlights first, then describe the light source in the opposite direction in your prompt. Color shift: The new background doesn't match the white of your other images. Fix: Use a color reference—save a white-balance sample from a successful image and reference it in your next prompt ("match background white from reference image provided"). Loss of product detail: The fill erased part of the actual item. Fix: Tighten your selection. Use the quick selection tool, then refine edges inward by 2–3 pixels before filling. Each failure gets a diagnostic flowchart: what to look for, what caused it, and the exact fix.
If you're processing 12+ images, structure matters. Step 1: Organize by shoot. Photos taken under the same lighting on the same day get one prompt approach. Photos from different days get individual adjustments. Step 2: Create a master template. Pick your best-looking image as a reference for color, shadow direction, and background tone. Screenshot the final prompt used. Step 3: Use an 8-point QA checklist after each fill. Check shadow consistency, background white-level matching, edge quality, reflection accuracy, color preservation, lighting direction, and overall polish. Rate each pass/fail and log what was fixed. Step 4: Batch export with consistent settings (sRGB, 72dpi for web, consistent file naming by product ID). This checklist prevents drift. By photo 8, your prompts stay consistent without you having to remember what worked on photo 2.
Generative Fill excels at: bad backgrounds, uneven lighting, perspective that's almost right, missing shadows, inconsistency across similar products, and rescuing phone-camera shots. Don't use it for: product defects (scratches, damage), wrong product angles, or color that's fundamentally off due to camera white-balance. Reshoot those. Generative Fill makes good photos great, not broken photos sellable. If 80% of your shot is salvageable, Generative Fill is worth 5–10 minutes. If 40% is usable, reshoot.