How to turn a photo into an AI prompt

A photo records a moment, not a complete specification.

Photographs feel factual, but they still leave important information unknown. A dark background may be a black wall, an underexposed room, or a distant space outside the focus plane. Warm light may come from sunset, a tungsten lamp, or color grading. A person’s expression is visible; their identity, occupation, and inner state usually are not.

A photo-to-prompt workflow should preserve the visible decisions while marking uncertainty. This produces a prompt that is accurate enough to recreate the image and flexible enough to edit.

This page focuses on photographic evidence. For illustrations and mixed references, use the broader image-to-text prompt workflow.

Choose one job for the reference photo.

A close recreation needs more observed detail. A style transfer needs fewer object descriptions and more explanation of light, color, surface, and photographic finish.

Separate observed facts from useful estimates.

Some photographic terms are estimates. You can often recognize a shallow focus plane or wide field of view, but you usually cannot prove an exact lens, aperture, shutter speed, or camera body from the pixels alone. Use the visual effect unless metadata confirms the setting.

• Identity: describe visible appearance; do not invent a real person’s name.
• Sensitive traits: do not infer health, religion, ethnicity, politics, or other private attributes from appearance.
• Brand and location: use generic descriptions unless a logo, sign, landmark, or reliable context makes the label necessary.
• Camera settings: describe the effect—wide field of view, compressed perspective, motion blur, shallow focus—unless metadata is available.
• Mood: connect it to evidence such as color, gesture, light, and space instead of presenting an interpretation as fact.

Use a six-pass photo reading.

1. Subject. Record count, visible features, clothing or material, pose, gesture, and action. Begin with the focal subject rather than the background.
2. Environment. Name the setting and only the supporting objects that establish use, scale, or story.
3. Composition. Identify orientation, shot size, viewpoint, subject placement, crop, leading lines, depth layers, and negative space.
4. Light and color. Locate the apparent source, direction, softness, contrast, highlight behavior, shadow density, dominant palette, and accent colors.
5. Photographic rendering. Describe depth of field, motion, grain, sharpness distribution, dynamic range, color grade, and whether the image feels documentary, editorial, commercial, or casual.
6. Constraints. For recreation, name what must stay. For edits, separate the requested change from the preserve list.

For portraits, protect pose and light before details.

A portrait can lose its character even when clothing and facial features are described accurately. Camera height, head angle, eye line, shoulder direction, crop, background distance, and catchlight placement often matter more.

This prompt does not claim identity or an exact lens. It preserves the visible portrait system: body direction, eye line, crop, light side, background distance, and finish.

When style is the main target, translate broad labels using the image prompt style guide.

For products, describe material behavior.

A product photograph depends on how surfaces react to light. “Metal bottle” is not enough. Brushed aluminum, polished chrome, anodized color, translucent plastic, frosted glass, matte paper, and glossy enamel need different highlight and shadow language.

• Name the surface finish, not only the material category.
• Describe which edge or texture the light should reveal.
• State camera height and product orientation when shape matters.
• Use negative space intentionally for labels, headlines, or a marketplace crop.

For travel and places, preserve spatial evidence.

Travel photos tempt writers to guess a location. A prompt often works better when it describes the architecture, terrain, weather, and human scale that are actually visible. If the exact place is important, verify it outside the image before adding the name.

The second version remains useful even if the location label is wrong. Architecture, slope, view corridor, light direction, and seasonal atmosphere are all visible controls.

Worked example: change the subject, keep the photograph.

Imagine a reference showing a ceramic vase on a stone plinth against a cool blue studio wall. The goal is to replace the vase with a perfume bottle while preserving the art direction.

Step 1: extract the stable setup

Step 2: add the new subject and its material needs

The composition and atmosphere stay stable, while the light instruction changes enough to support transparent glass. That is a deliberate transfer, not a global “make this with another product” request.

Run a final evidence check.

• Did the prompt separate visible facts from guesses?
• Is the intended task—recreate, replace, restyle, or edit—obvious?
• Do shot size, camera height, and crop describe one possible frame?
• Does the light interact plausibly with the stated materials?
• Are color and mood connected to visible evidence?
• For an edit, is the requested change separate from the preserve list?
• Have accidental compression, blur, clutter, or text been excluded only when appropriate?

Compare your draft with the annotated prompt examples, or turn your photo into a prompt and use this checklist as the editing pass.