// free tool · runs in your browser · no signup
Nano Banana Prompt Checker
frame it before you generate
Paste an image prompt, see what the official guide would change, copy the fixed version. Also for GPT Image 2.5.
Prompt layers
Frame the prompt before you generate
6 checks from the official guide · paste a prompt or
Nano Banana vs GPT Image 2.5: where the guides differ
From Google's Nano Banana prompting guide and OpenAI's GPT Image 2.5 prompting guide, checked on 2026-10-25:
| Nano Banana | GPT Image 2.5 | |
|---|---|---|
| Things to avoid | Describe the positive: "empty street", not "no cars" | "No watermarks"-style exclusions are fine |
| Prompt format | Narrated scene: subject, action, place, composition, style | Any format; labeled sections for complex prompts |
| Text in the image | Exact words in quotes + font + position | Exact words in quotes + position + "exactly once" |
| Editing | Say what changes and what stays the same | "Change only X" + list what to preserve |
| API settings | aspectRatio, imageSize | size, quality, background |
Frequently asked questions
What does this Nano Banana prompt checker check?
Six things from the official guides: "no …" phrasing (Nano Banana wants the positive description), text in the image without the exact words in quotes, aspect ratio, resolution or transparent background written in the text when you use the API, prompts written as a keyword list, edits that don't say what to keep, and multiple reference images without a role for each. Each finding cites the Google or OpenAI guide it comes from.
Nano Banana or GPT Image 2.5: which rules change?
Two big differences. Google's Nano Banana guide says to describe what you want instead of what you don't ("empty street" instead of "no cars") and to narrate the scene instead of listing keywords. OpenAI's GPT Image 2.5 guide uses "No watermarks"-style exclusions in its own examples and accepts any prompt format. Both ask for exact text in quotes, and both ask you to say what must stay the same when editing.
Should I write the aspect ratio in the prompt?
In the Gemini app or ChatGPT, yes: it is the only way to ask. In the API, no: set it as a parameter (aspectRatio and imageSize in the Gemini API, size and background in the OpenAI API). Pick the surface at the top and the checker applies the right rule. For a transparent background in the OpenAI API, use background="transparent"; asking in the text tends to draw a checkerboard.
How do I get readable text in an AI image?
Put the exact words in quotes, say where they go and in what font, and (for GPT Image) how many times they appear: Headline "SUMMER SALE" in bold white sans-serif, top center, rendered exactly once. The checker flags text requests without quotes and quotes the words for you when it can tell what they are.
Is it free? Do you store my prompt?
Free, no signup, no limit. The check runs entirely in your browser: the prompt is never sent to a server.
What about Midjourney?
Not yet. Midjourney handles exclusions with the --no parameter rather than in the text, and its default version changed to V8.2 in July 2026, so it needs its own rule set. Nano Banana and GPT Image 2.5 come first because both publish official prompting guides.