How to Write Effective Scene Prompts

How to write prompts that get great results from AI-generated scenes

Written By Bonnie FaxΓ©n

Last updated 2 months ago

1. What is a scene prompt?

A scene prompt describes the full visual context of an image β€” not just the garment, but the world around it. You're directing two things at once: the background (where the model is) and the pose (how the model looks and moves).

The golden rule is the same: be specific. The AI can't guess your creative vision β€” it needs you to describe it.

Too vague: A model outside wearing the jacket

Much better: A female model standing on a cobblestone street in a European city, soft overcast daylight, looking slightly off-camera to the left, relaxed posture with one hand in pocket

2. The 2 ingredients: background & pose

Background

Think of this as your set design. Describe the location, lighting, time of day, and mood.

Location

  • Urban: city street, rooftop, industrial warehouse, subway entrance

  • Nature: pine forest, open field, rocky coastline, snowy mountain

  • Interior: minimalist apartment, changing room, coffee shop, studio

  • Abstract: plain color backdrop, gradient, soft bokeh (blur quality)

Lighting

  • Natural: golden hour sunlight, overcast diffused light, harsh midday sun, blue-hour dusk

  • Studio: soft box lighting, rim lighting, high-key white, low-key moody

  • Atmospheric: fog, mist, rain-slicked surfaces, dappled forest light

Mood & color palette

  • "Muted, earthy tones β€” warm browns and dusty greens"

  • "Clean and minimal β€” white and cool grey"

  • "Cinematic and moody β€” deep shadows, desaturated color"

Pose

Think of this as your casting direction. Describe the model's body position, energy, and gaze.

Body position

  • Standing: "standing upright, weight on left leg, relaxed shoulders"

  • Walking: "mid-stride, natural movement, arms swinging slightly"

  • Seated: "sitting on a low wall, legs apart, leaning forward on elbows"

  • Leaning: "leaning against a wall, one shoulder back, hip tilted"

Energy & expression

  • Relaxed and candid: "looking down at phone, natural off-guard expression"

  • Editorial: "strong direct gaze at camera, chin slightly down, serious expression"

  • Movement: "turning to walk away, glancing back over shoulder"

  • Lifestyle: "laughing, mid-conversation, animated and natural"

Hands

Hands are easy to leave vague and easy for AI to get wrong. Be explicit:

  • "Both hands in jacket pockets"

  • "One hand adjusting collar, other arm loose at side"

  • "Carrying a tote bag in right hand"

  • "Arms crossed loosely at chest"

Gaze direction

  • Direct to camera / slightly off-camera left / looking down / looking into distance

3. Scene prompt templates

  • Urban editorial [Model description] standing on [urban location], [time of day] with [lighting quality]. [Pose detail]. [Gaze direction]. Color palette: [mood/tones]. Full-length shot. Example: A tall female model standing on a wet cobblestone street at dusk, warm streetlights reflecting off the pavement. Weight on her right leg, left hand in coat pocket, right arm loose. Looking slightly off-camera to the left. Color palette: warm amber and deep shadow. Full-length shot.

  • Nature / outdoor lifestyle [Model description] in [natural setting], [time of day], [weather/light]. [Pose β€” movement or stillness]. [Gaze]. Background: [depth and detail of scene].Example: A male model walking through a pine forest in early morning light, mist low between the trees. Mid-stride, arms relaxed, looking ahead into the forest. Background softly out of focus with depth of field.

  • Studio / clean background [Model description] against a [color/texture] studio background, [lighting setup]. [Pose detail]. [Expression/gaze]. [Shot framing β€” full length / half body / portrait].Example: A female model against a warm off-white studio background, soft diffused side lighting. Standing with weight shifted left, one hand lightly touching her collar, relaxed expression, direct gaze to camera. Half-body shot.

  • Seated / lifestyle [Model description] seated on [surface/location], [setting detail]. [Pose]. [Expression β€” natural and candid or directed]. [Framing]. Example: A male model seated on the steps of a concrete building, urban setting, overcast daylight. Leaning forward with elbows on knees, hands loosely clasped. Natural candid expression, looking slightly downward. Three-quarter shot.

4. Getting scenes to look consistent

Consistency across a shoot is harder with scenes than with white backgrounds β€” but these habits help a lot.

  • Lock your lighting description and reuse it word for word across all shots in a collection. Even small changes ("golden hour" vs "warm afternoon sun") can shift the mood significantly.

  • Define your model description once β€” height, build, hair, skin tone β€” and paste it into every prompt in the session.

  • Keep background depth consistent β€” decide early whether you want a sharp background or bokeh, and state it explicitly in every prompt.

  • Vary the pose, not the setting β€” if you're shooting multiple garments in the same scene, keep the location and lighting identical and only change the garment and pose details.

  • Generate in one session β€” the AI holds more context within a single session, so front and back views or multiple looks from the same scene will stay more coherent.

5. When the scene isn't working

The background is too busy / distracting

  • Add "background softly out of focus" or "shallow depth of field"

  • Simplify: "plain [color] wall" or "minimal setting, attention on the model"

The pose looks stiff or unnatural

  • Add movement language: "mid-step", "shifting weight", "in the middle of turning"

  • Describe energy, not just position: "relaxed and candid" or "confident and still"

  • Be explicit about hands β€” unspecified hands often look awkward

The lighting doesn't match the garment

  • Describe how you want light to fall on the fabric: "soft light catching the texture of the jacket", "even lighting to show colour accurately"

  • For dark garments, avoid dark moody scenes β€” specify "well-lit" even in atmospheric settings

The model's gaze breaks the mood

  • Direct gaze is strong and editorial β€” use it intentionally

  • Off-camera or downward gaze reads as more candid and lifestyle

  • State it explicitly: the AI defaults inconsistently if you leave it open