FRI, OCTOBER 09, 2026
Independent · In‑Depth · Practitioner‑Tested
Gemini Image

Gemini Image Prompts That Work on Nano Banana 2.1

Nano Banana 2.1 went generally available on 6 October at half the per-image price of its predecessor, and the 512px tier is gone — so every render now starts at 1K. These eight prompts are built for that: they specify composition, lighting and output intent explicitly rather than relying on a long adjective pile, which is what actually separates a usable render from a retry. Use them as a reference set when you test 2.1 against the model you are migrating from.

⌨️ 8 prompts 🕐 Updated Oct 9, 2026
💡 How to use these prompts: Replace everything in [BRACKETS] with your specific details before sending. Click Copy to copy any prompt to your clipboard instantly.
1
Clean product shot
E-commerce and catalogue images
A single [PRODUCT] centred on a seamless light grey backdrop. Soft key light from the upper left, a weak fill from the right, and a narrow specular highlight along the top edge. No props, no hands, no text. Shallow depth of field with the front face fully sharp. Square framing with even margins on all four sides. Photographic, neutral white balance, no stylisation.
2
Hero image with space for text
Blog headers and landing pages
A wide landscape scene of [SUBJECT], composed so the left third is visually quiet and uncluttered for a headline to sit over it. Subject weighted to the right. Low contrast in the quiet area, higher contrast on the subject. Cool colour grade. No text in the image. Render at the widest aspect available and keep the horizon level.
3
Technical diagram
Documentation and explainers
A clean flat diagram showing [PROCESS] as [N] labelled stages left to right, connected by simple arrows. Flat colours, no gradients, no drop shadows, no three dimensional effects. Generous white space. Each stage is a plain rounded rectangle of the same size. Use one accent colour and grey for everything else. Legible at half size.
4
Consistent character across images
Building a reusable character reference
Create [CHARACTER DESCRIPTION INCLUDING AGE RANGE, BUILD, HAIR, CLOTHING AND TWO FIXED DISTINGUISHING FEATURES]. Full body, neutral standing pose, plain background, even lighting, no expression. This is a reference sheet. Describe the result back to me in one paragraph of specific visual detail that I can paste into later prompts to keep the character consistent.
5
Editing an existing image
Targeted edits that do not drift
Using the image I have attached, change only [SPECIFIC ELEMENT]. Keep the composition, lighting direction, colour grade, grain and every other element exactly as they are. Do not recrop, do not shift the subject, and do not alter anything outside the element named. If the change cannot be made without affecting the rest, tell me what would be affected before you render.
6
Icon or avatar at small size
Favicons, app icons and avatars
A simple symbolic mark representing [CONCEPT]. Bold shapes, one accent colour on a flat background, heavy line weight, high contrast. It must stay readable at thirty two pixels, so no fine detail, no text, no thin lines, and no more than three distinct shapes. Centred with even padding. Render at 1K and I will downscale.
7
Scene with specific lighting
Controlling light instead of hoping
[SUBJECT] in [SETTING], lit by [LIGHT SOURCE] from [DIRECTION]. Describe the light rather than the mood: name the key, the fill ratio, where the shadows fall and how hard their edges are. Realistic falloff. No lens flare, no bloom, no artificial vignette. Keep the colour temperature consistent across the whole frame.
8
Reference set for model comparison
Testing one model against another
Render this exact prompt with no interpretation or embellishment: [PROMPT]. Do not add elements I did not ask for. Do not improve the composition. I am using this output to compare two models, so I need the literal reading of the text rather than the best possible image. Note at the end anything in the prompt you found ambiguous.