How to Write AI Prompts That Produce Real Product Photos

Most AI product shots come out looking almost right and still unusable: the light is flat, the angle is odd, or the product looks like plastic. The fix is rarely a better model. It is a prompt that names the five things a real photographer would decide before pressing the shutter. Get those five right and the same prompt keeps working, product after product.

Name the five things a real shoot decides

A vague prompt like "product photo of a candle, professional" leaves the model to guess everything, so it averages toward stock-photo mush. A working prompt makes five decisions on purpose: the product, the lighting, the camera angle, the lens, and the frame. Think of it as a shot brief. For example: "amber glass candle on a pale travertine surface, soft window light from the left, three-quarter angle slightly above, shot on a 50mm lens, shallow depth of field, warm neutral tones." Every clause there replaces a guess with an instruction. Start every prompt from these five slots and fill each one, even briefly. The difference between a throwaway image and one you can put on a listing is almost always one of these five being left blank.

Describe the light, because light is the photo

Lighting is what makes a shot read as real or as rendered. The flat, evenly lit look that AI defaults to is the giveaway, so name a direction and a quality instead. Direction: "light from the left," "backlit," "top-down." Quality: "soft diffused window light" for a calm editorial feel, "hard directional light with sharp shadows" for drama, "golden hour" for warm lifestyle scenes. Add a modifier for mood if you want it, like "long soft shadows" or "gentle reflection on the surface." Soft window light from a 45-degree angle flatters almost any product and is a safe default. Avoid asking for "studio lighting" alone, which the model reads as flat and lifeless. One clear light source described well beats three vague ones.

Set the angle and lens for the product you have

Angle and lens control how the product sits in space, and different products want different treatment. A three-quarter angle slightly above eye level suits most objects, since it shows the front and the top at once. Flat-lay, meaning straight top-down, works for anything arranged on a surface: stationery, jewelry, food. A low, near-eye-level hero angle makes bottles and packaging look tall and important. For the lens, "50mm" or "85mm" reads natural and close to how a phone or portrait camera sees, while "macro" pulls in tight on texture and detail. Add "shallow depth of field" to blur the background and separate the product, or "deep focus" when the whole scene needs to stay sharp. Match the angle to the product's best face, not to a habit.

Pick the aspect ratio before you generate

Aspect ratio is the cheapest quality decision and the one most people skip. Generating square and then cropping to vertical wastes half the frame and often cuts the product badly, so set the shape up front to match where the image will live. Square, written as 1:1, is the default for most marketplace and grid thumbnails. Vertical 4:5 is the strongest ratio for Instagram feed and Pinterest, since it takes up more screen. Tall 9:16 is for stories and reels. Wide 16:9 or 3:2 suits website banners and headers. In gpt-image and Midjourney you state this directly, for instance "--ar 4:5" in Midjourney or by asking for a portrait image. Decide the destination first, set the ratio to it, and compose the scene with room around the product so nothing important sits at the edge.

Turn the winner into a reusable [product] template

Once a prompt gives you a shot you like, do not retype it for the next item. Freeze it into a template with the product swapped for a placeholder, like this: "[product] on a pale travertine surface, soft window light from the left, three-quarter angle, 50mm, shallow depth of field, warm neutral tones, 4:5." Now you replace [product] with a clear description of each item, keeping the noun specific: "a matte black ceramic mug" reads far better than "mug." Keep the scene and light fixed so your shots share one consistent look across a whole range, which is what makes a shop feel considered rather than random. Build a small set of these templates, one per scene type, flat-lay, lifestyle, hero, and you have a repeatable system instead of a lucky roll each time.