What this GPT Image-2 generator optimizes
- Lens Look instead of numeric focal length and aperture
- Output size mapped from your aspect ratio (e.g. 2:3 → 1024×1536)
- Quality, background, format and compression shown as settings, not prompt text
- Things to avoid written as plain constraints
How GPT Image-2 prompts are built
Order: subject, appearance, clothing, pose, scene, background, lighting, framing, camera language, style, references, constraints. No --ar, --raw or --no.
Example GPT Image-2 prompt
A studio portrait of a woman in her twenties, head-and-shoulders framing, photographed at eye level with a flattering portrait lens look using a very shallow depth of field.
Product / Fashion with references
Product reference + clothing reference, a model and editorial style — the product stays the visual hero.
Create high-fashion editorial photograph with bold composition and styling of a tailored cream linen blazer worn by a female model, the outfit is the hero of the frame with sharp focus on the clothing, showing the garment's silhouette, fabric texture, drape and construction clearly, while the background stays subtle and non-distracting. Medium shot framed approximately from the waist up. Use the reference image to use the product shown as the featured object, preserving its recognizable design, proportions, colors, materials, and distinctive details; do not copy the reference person's identity, pose, background, lighting and camera setup unless separately requested. Compose the image as a square composition, 1:1 aspect ratio.
Multi-reference: identity, pose and environment
Shows how GPT Image-2 maps each reference to its role. You attach the images in GPT Image-2.
Create hyper-realistic photograph with lifelike detail and accurate textures of a woman in her thirties. Full-body image showing the subject completely from head to toe. Use the reference image to preserve the subject's recognizable facial identity and facial features and match the body pose while preserving the independently specified subject appearance and scene; do not copy facial expression, background, lighting and camera setup unless separately requested. For image 2: use the environment, setting and spatial characteristics from the reference image while allowing the subject and other elements to change. Compose the image as a square composition, 1:1 aspect ratio.
GPT Image-2 FAQ
- Where do quality and size go?
- They are API/app settings. PhotonPrompt shows them in a separate Model settings box so the prompt stays clean.
- Why no numeric aperture?
- GPT Image-2 responds to described looks better than camera numbers, so depth of field is described in words.