Best PromptsPhoto Editing Multi-Tool
The Keep-Change-Match formula behind Nano Banana's viral edits: background swaps, object removal, and relighting that preserve everything else.
[Attach your photo, then build the edit from this formula:] KEEP: state what must not change, explicitly: "Keep [the subject's face, expression, skin tone, hair, clothing, and pose / the product exactly as it is / everything except what I name below] completely unchanged." CHANGE: one edit per generation: For a background swap: "Replace the background with [NEW SCENE, describe it like a location brief: 'a quiet beach at golden hour, soft waves out of focus behind']." For object removal: "Remove [THE OBJECT, be locational: 'the man in the red jacket on the left edge' / 'the power lines crossing the sky'] and fill the space naturally with the surrounding [sand/sky/wall]." For relighting: "Change the lighting to [THE NEW LIGHT: 'warm golden-hour sunlight from camera left' / 'soft overcast daylight' / 'moody window light from the right'], adjusting the mood while keeping every object in place." MATCH: make the physics agree (the line most edits miss): "Match the lighting direction and color on the subject to the new [background/light source] (shadows falling [direction], [warm/cool] tones in the highlights) so the result looks like one real photograph, not a composite." Then iterate conversationally with single instructions: "same image, but soften the shadow under the chin" / "keep everything, make the background 20% more blurred".
How to use
Editing is what made Nano Banana the breakout image model, and this Keep-Change-Match structure is why some edits look real and others look pasted: the Keep list stops identity drift, one change per generation keeps the model precise, and the Match line makes the lighting physics agree. Chain edits across turns in the same conversation rather than writing one mega-instruction: the model holds context between edits.
More image prompts
[Attach the reference image, then use this prompt:] Analyze this image and write me a prompt that would reproduce its look with a different subject. First, describe what you actually see, in these terms: - **Subject and composition**: what is in frame, where it sits, how much of the frame it fills, the crop and aspect ratio, the eye lev
Reverse Engineer an Image Prompt
Turn any reference image into a reusable prompt by naming its subject, lighting, lens, palette, and the one detail that carries the look.
Build me a reusable image style kit for [BRAND]. Brand personality in three words: [e.g. "warm, precise, unfussy"] Who it speaks to: [AUDIENCE] Brand colors: [HEX VALUES, or "you propose"] References I like: [DESCRIBE THEM, OR ATTACH THEM] What we must never look like: [THE COMPETITOR OR AESTHETIC TO AVOID] Where these images go: [WEBSIT
Brand Image Style Kit
Define one reusable visual formula so every image you generate looks like it came from the same brand, across subjects and use cases.
Create a t-shirt design. Concept: [THE IDEA, JOKE, PHRASE, OR IMAGE] Exact text, if any: [TYPE IT EXACTLY, WITH THE CAPITALIZATION YOU WANT] Audience and where they would wear it: [WHO BUYS THIS] Style: [bold vintage / minimal line art / retro 70s sunset / hand-drawn / heavy blackwork / kawaii] Shirt color it prints on: [DARK / LIGHT / B
Print on Demand Apparel
Generate t-shirt and merch artwork built for printing: flat color, clean edges, transparent background, and a mockup that sells it.