ChatGPT photo prompts: write an edit brief that preserves the subject

Separate a new scene from an edit to an existing photo

A text-only scene brief and an edit to a supplied photograph solve different problems. A scene can invent a subject and setting. An edit should explain which parts of the source must remain recognizable. Begin by stating which job you want, then name the subject, change and output use. “Make it professional” does not establish whether clothing, facial appearance or the room may change.

For a product photo, write down color, shape, label text and proportions before changing the background. For a portrait, identify the permitted adjustments and the person’s consent. Keep identity-sensitive corrections narrow and review the result against the reference rather than treating resemblance as automatic.

Use official prompting guidance without mixing product limits

OpenAI’s image-prompting documentation explains reference-based edits and specific visual constraints. Its API examples are useful for writing a brief, but they do not establish the allowance or interface of a particular ChatGPT subscription. The official image-input guide explains supplying an image and describing what to inspect. Confirm image-generation access in the account you will actually use.

Choose a conversational image workflow when you want to describe a scene and refine a small change. Choose a conventional editor when you need precise pixel selection, repeatable color management or a layered production file. Current ChatGPT image quotas, download options and the price of the intended account plan were not verified for this article.

Worked example: preserve a ceramic mug while changing its setting

Completed fictional editorial brief: a matte blue mug with one curved handle and no logo, photographed on a table. Initial prompt: “Make a luxury coffee advertisement.” That request could introduce gold trim, change the handle or add a brand name. We revised the brief before generation.

Final prompt: “Use the supplied mug photograph as the subject reference. Keep the matte blue finish, one curved handle, proportions and absence of a logo. Replace only the surroundings with a pale stone counter and soft window light from the left. Show the entire mug, leave open space on the right for separately typeset copy, and add no text or additional products.” This is a completed prompt-writing exercise. No image was generated or judged in ChatGPT for this guide; its effect must be checked with the actual output.

Revise one failure at a time

If the resulting handle changes, ask for that correction while preserving the accepted lighting and composition. If the copy area is crowded, describe the required empty region instead of requesting another vague redesign. Keep a record of the accepted frame so a later correction does not quietly undo earlier improvements.

Review object count, subject geometry, shadows, readable labels and crop. For marketing, place final text in an editor where it can be checked and changed. A prompt requesting accurate lettering is an instruction, not a guarantee that every letter will survive rendering.

A reusable prompt card and a practical handoff

Keep five fields on the prompt card: source image, preserved facts, permitted change, composition and rejection criteria. A short explicit card is easier to revise than a long collection of style adjectives. Export only after reviewing the actual file at the size where it will be used. Record the account, date and any generation limit encountered.

For planning, if six attempts produce two accepted frames, the acceptance rate is two out of six. That fictional calculation describes review efficiency, not vendor performance. Do not turn attempts into cost without the actual plan’s billing units. Save the original photograph and final prompt with the accepted asset.

Image prompts questions