How do I write a good prompt for AI product photos?

A good product photo prompt is a shot description a photographer could act on. Most weak prompts fail the same way: they describe the outcome they want (“professional, high quality, stunning”) instead of the physical scene that produces it. The model cannot use your ambition. It can use a surface, a light source and an angle.

The five-part structure

Build the prompt in this order and you will rarely go wrong:

1. Subject. What the product is, its material and colour: “a matt white ceramic candle jar with a wooden lid”.
2. Surface. What it sits on, because the surface sets half the mood: “on a honed travertine slab”, “on a weathered oak table”, “on wet black slate”.
3. Light. Direction and quality, the single highest-leverage line: “soft window light from the left with gentle shadows”, “hard noon sun from above”, “one warm spotlight in a dark room”.
4. Lens and angle. One phrase for framing: “85mm close-up, shallow depth of field” for a tight hero shot, “35mm at eye level” for an environmental scene, “overhead flat lay” for a top-down.
5. Mood. One or two words at most: “calm, premium”, “fresh, energetic”. Mood words fine-tune; they cannot carry a prompt alone.

Put together: “a matt white ceramic candle jar with a wooden lid, on a weathered oak table beside a small stack of linen, soft window light from the right, 50mm at eye level, quiet Sunday-morning mood.”

Then subtract. Remove any word that does not change the picture, and never stack contradictions like “bright airy dramatic moody”. One scene per prompt; if you want two looks, run two prompts.

The honest limits of prompting

Two things a great prompt cannot fix. It cannot make a generator render your actual product; a text description always produces a lookalike with an invented label. And it cannot replace taste: you will still iterate, nudging light and props over several attempts.

That iteration is exactly what scene-based tools remove. BrandStills skips the prompt entirely: each scene in the library is a prewritten photographic setup, surface, props, light direction and lens already specified the way the structure above specifies them, from a clean white marketplace shot to a hard-shadow studio look. You photograph your real product two or three times with your phone, pick the scene, and the render keeps your label, colours and proportions locked. You can still refine in plain words afterwards (“warmer light”, “move it left”), with each edit building on the previous image, and you can paste an inspiration image instead of describing one.

Learn the five-part structure; it makes you better at judging any product image. But if you would rather choose a finished setup than engineer one, get the app and let the scene do the prompting.