AI photo-editing prompt recipe

Change photo lighting with a bounded text prompt

To change photo lighting with a text prompt, name the lighting you want and the scene it should affect, then state what must stay recognizable. Start with one bounded change, such as “make it a pink and orange sunrise,” and review the subject, edges, shadows, color spill, text, and composition across the complete result before accepting it.

Published source and result

The exact source/result pair below is already published by Peelaway with the prompt shown here. Compare the complete frame, not only the most convincing changed area.

Source photograph used in Peelaway's published sunrise-lighting example
Source image
Published result paired with the prompt Make it a pink and orange sunrise
Published result

Exact published prompt

Make it a pink and orange sunrise

Evidence boundary: This is a published example, not a customer case study. Peelaway is not asserting run provenance for this pair. Output varies with the source and requested change, so inspect the complete result before using it.

Reusable prompt template

Make the requested change concrete, then name the details that should remain recognizable. A preserve clause guides review; it is not a guarantee that pixels outside the target stay unchanged.

Change [lighting or time of day] to [desired light]. Keep [subject, composition, and important details] recognizable. Do not change [protected elements].

Two bounded variations

  1. 01Shift the scene to soft overcast daylight. Keep the subject, framing, and building shapes recognizable.
  2. 02Warm the late-afternoon light without changing the subject, composition, or visible text.

These are adaptable wording examples, not outputs from the published source/result pair.

Review the complete result

  • Compare the main subject's identity, shape, and position.
  • Inspect boundaries where the subject meets the sky or background.
  • Check whether highlights and shadows follow the requested light direction.
  • Look for unintended color spill on neutral or protected objects.
  • Read every visible word, logo, and number again.
  • Review the full frame rather than accepting a convincing focal area alone.

Test the wording on an image you can review

Start with a representative source, keep the request bounded, and compare the complete output with the acceptance threshold for your use case.