A photoshoot prompt that works names five things in this order: who is in the photo, where they are, what they wear, how the light falls, and one line that tells the model to keep the face exactly the same. Every photoshoot tool on Photoshoot.si is built on that pattern, and each look a tool offers is one filled-in version of it, so you can borrow the structure for your own prompts.

The five-part pattern

Here is the prompt behind the sample at the top of this page, made with the AI Photoshoot tool, with the parts labelled:

  1. Subject. "Professional photoshoot of the person in the reference photos". This tells the model who to draw and where to look for them.
  2. Setting. "golden hour on a city street". One place, one time of day.
  3. Outfit. "black blazer over a white tee". Name the garments, not the vibe.
  4. Camera and light. "Shot on a full-frame camera, 85mm lens, natural skin texture". Camera words push the model toward a photograph instead of an illustration.
  5. Keep line. "Keep their face, identity and hair exactly the same." Without it, image models tend to smooth or change features.

The sample came from exactly that prompt and one reference photo.

Looks: one prompt, a whole shoot

A real photoshoot is a set of photos, not one frame. On Photoshoot.si the tools that suit a set come with looks: each look is a complete setting, outfit and light description that fills the setting part of the pattern. Pick one look for a single image, or press Full shoot to generate one image per look with the same photos and model. The subject, camera and keep lines stay the same across the set, which is why the frames look like they came from one day.

Good looks pair a place with a light source and an outfit, for example:

  • Clean white studio backdrop, soft key light, black turtleneck
  • Golden hour on a city street, denim jacket
  • Beach at sunset, linen shirt
  • Minimal apartment with window light, oversized knit
  • Rooftop at night with city lights, leather jacket

"Beach" alone leaves the model to guess the time of day; "beach at sunset" settles it. When you write your own setting, keep that pairing.

Concepts beat adjectives

The AI Generated Photoshoot tool works from a concept instead of a setting, the way a real shoot starts from a moodboard. Concepts like these carry styling, props and light in a few words:

  • Old money summer on a yacht
  • 90s film camera flash party
  • Parisian cafe, morning light
  • Cozy autumn knitwear in a forest
  • Streetwear in Tokyo at night
  • Black and white fine art portrait

That is why a concept reads as planned. A details field adds one or two specifics on top: "wearing a cream linen suit, laughing" beats "stylish and happy" because the model can draw a linen suit and a laugh; it cannot draw "stylish".

Building a set with shot type and grade

A set looks like one shoot when three things stay fixed across it: the theme, the light and the color grade. The AI Photoshoot Generator splits those into separate fields so you can change one at a time: a free-text theme, a shot type (close-up, waist-up, full body, candid, over the shoulder) and a color grade (natural, warm film, cool cinematic, black and white, soft pastel).

Keep the theme and grade the same and vary the framing. The result is a varied set that still matches, which is what a real photographer would hand you.

Where your words go on a tool page

On a tool page you do not write the whole prompt. The tool fills in its template from the look and fields you choose, and your typing only goes into the free-text fields (outfit, details, theme). Open "Prompt sent to the model" under the fields to read the exact text before you generate.

If you want to change the structure itself, for example to drop the camera line or add a second person, open the studio, choose Edit photo and write the prompt from scratch. The pattern above still applies.

Mistakes that cost a generation

  • Stacked adjectives. "Stunning, beautiful, ultra-realistic, 8k" adds nothing the model can draw. Replace them with a light source or a lens.
  • Two settings in one prompt. "Beach and city rooftop" gives you a confused background. Make them two looks instead.
  • Conflicting light. "Soft window light, harsh noon sun" makes the model pick one or blend both badly.
  • Dropping the keep line. It is the single line most responsible for the result still looking like you.
  • Blocked words. Prompts are checked before any credits are spent. Nudity words are always blocked when a photo is attached, even in an innocent context. See why a generation was blocked.

Note

Even a good prompt is a request, not a guarantee. The model is drawing a new picture, so some frames will look more like you than others. Generate a single look first, and run the Full shoot once the likeness is right.

Picking a model for your prompt

The same prompt reads differently on different models. Nano Banana is the default on most photoshoot tools because it keeps faces consistent. It accepts up to three reference photos per run.

Seedream 5.0 Lite is the default on the AI Photoshoot Generator. It can return up to four images per run, which suits building a set from one prompt, and it tends toward more cinematic color.

A practical test: run your best look on both, with the same photos, and compare likeness first and styling second. The Generate button shows the credit price for each model before you spend anything, and a failed job is refunded automatically.