Listing images and clean studio shots.
Marketplace white background
[product] centred on a seamless pure white background, even diffused studio lighting, no harsh shadows, sharp focus across the whole product, straight-on eye-level angle, generous margin around the edges
Marketplaces reject shots that crop tight or cast strong shadows. Asking for even diffused light and margin up front avoids a re-run.
Soft studio with contact shadow
[product] on a matte [surface colour] surface, single large softbox from the upper left, soft contact shadow beneath the product, subtle falloff to darker background, shallow depth of field
A contact shadow is what stops a product looking pasted on. Naming one light source keeps the result readable rather than flat.
Lifestyle in context
[product] on a [surface] in a [room type], [time of day] light through a window from the side, a few out-of-focus everyday objects in the background, natural and unstaged
Specify the light direction and time of day. "Natural lighting" alone is the vaguest instruction in product prompting.
Detail and texture macro
extreme close-up of [specific part] of a [product], raking light across the surface to reveal [material] texture, fine detail, shallow depth of field with the background falling away
Raking light - grazing across the surface rather than pointing at it - is what makes texture read. Name the specific part, not the whole object.
Specific beats elaborate
All 24 prompts here follow the same principle: name the subject, then constrain the things that actually decide whether the image works - light direction, angle, background, and what to leave empty. Long prompts stuffed with quality adjectives tend to do less than short ones that state a clear constraint.
Say what you want left out
Half of prompting well is exclusion. Plain background, no text, no dramatic shadows, nothing cropped - these are the instructions that make an output usable downstream, especially if the next step is background removal or 3D reconstruction, both of which fail on busy or cropped inputs.
Reserve space deliberately
If text, a logo or platform UI is going over the image later, the prompt should reserve that room up front. Generating a full composition and then covering a third of it is the most common avoidable mistake in social and ad work.
Iterate cheaply, then commit
Because image generation is the cheapest thing on the platform, the sensible loop is to refine a prompt at the image stage and only then spend on video, 3D or upscaling. Getting the prompt right before the expensive step is where the savings actually are.
Frequently asked questions
- Do I use these exactly as written?
- No - the bracketed slots are the point. Swap in your product, colour, theme or character and adjust from there. A prompt you have edited to match your subject will always beat one pasted verbatim.
- Why does each prompt have a note?
- Because the technique transfers and the wording does not. Knowing that raking light reveals texture, or that reel symbols must read at small sizes, lets you write the next prompt yourself instead of hunting for a template.
- Will the same prompt give the same image twice?
- No. Generation is not deterministic, so expect variation between runs of an identical prompt. That is useful - run a prompt a few times and pick, rather than judging it on one result.
- Do these work across different models?
- The structure carries over, but models respond differently in emphasis and style. Treat a prompt as a starting point when you switch models rather than assuming identical behaviour.
Related tools and workflows
Ready to create with AI?
Start with free trial credits. Pay only for the generations you run.
Try a Prompt Free →