Last month I needed twelve product photos for a client’s Shopify listing and had exactly zero dollars left in the shoot budget. The client had one decent photo of the item on a plain table, taken with a phone, mediocre lighting. My job was to turn that into a dozen usable lifestyle shots by Friday, so with no camera and no budget, I opened eight browser tabs and started running that one photo through every AI photo generator I could find.
Two years ago I tried an AI photo generator for a different project and the results looked like wax figures with extra fingers, so I wasn’t expecting much going into this. The tech had presumably improved since then, but “improved” and “good enough that a paying client won’t notice” are different bars, and I genuinely didn’t know which side of that line I’d land on.
The first three tools I tried were free, and it showed. One kept adding a faint watermark I couldn’t remove without paying. Another nailed the lighting but completely reinvented the product every time, changing the color, the shape, even the label text, which is a fun way to accidentally advertise a product that doesn’t exist. The third was fine for a background swap and nothing else — ask it to add a hand holding the product and you’d get six fingers, every time, like it was daring me to notice.
By tool four I’d started keeping notes, because at that point I couldn’t remember which one had done what to which photo anymore. The pattern: cheap tools were good at exactly one job, a background swap or an upscale or a filter, and broke the moment I asked for more than that. What I needed was a tool that wouldn’t quietly turn my client’s actual product into a slightly different, imaginary product every time it generated a new scene.
Artlist’s AI photo generator is the one that ended up doing most of the actual work, mostly because of the image-to-image mode. I could upload the client’s original photo as a reference instead of typing a prompt and hoping for the best, so the product itself didn’t drift between the twelve shots the way it had with the earlier tools. There’s a stack of different models to choose from too — one leaned punchier and more commercial with the light, another matched the product’s real colors more closely — and switching between them didn’t cost extra or send me back to square one.
What actually saved my Friday deadline, though, was that it handled multiple aspect ratios in a single pass: square for the listing, vertical for a Pinterest pin, wide for the banner at the top of the page. No regenerating from scratch three separate times. One run through the AI photo generator per background scene, in every ratio I needed, and all twelve images were done in under two hours — most of which I spent staring at lighting presets trying to decide which one looked least fake.
Reflective surfaces are still the tell — if your product has any glass or chrome on it, expect to run a few extra passes and eyeball the reflections carefully before you send anything to a client. I’d also say: don’t skip the step of comparing the output side by side with the real product on your own screen at full size. Small stuff, like a logo getting slightly warped, is easy to miss when you’re scrolling past a thumbnail.
What changed for me isn’t that I now think every image can be faked convincingly — plenty still can’t. It’s that the gap between “obviously AI” and “would pass at a glance in a product listing” has gotten a lot smaller than it was the last time I tried this. Two years ago I would have told you an AI photo generator wasn’t worth the time for anything client-facing. Now it’s the first thing I reach for when a client hands me one mediocre photo and an unreasonable deadline.
If you’re doing this kind of work regularly, it’s worth setting aside an hour to actually test a few tools on your own product photos rather than trusting a demo reel — the demo shots are always the best-case examples, and your actual product, with its actual weird reflections and actual imperfect lighting, is the real test.
