What’s new
- Its official developer/API name is Gemini 3.1 Flash Image: Nano Banana 2 is what it's called in the Gemini app and other consumer products; Google positions it as "Pro-level image generation and editing, Flash-level speed."
- Keeps up to 5 characters and 14 objects consistent: it maintains people's likeness and object details across a single workflow, good for a series of images that need the same character or product to keep reappearing.
- Pulls in real-world information before drawing: Google says it draws on Gemini's knowledge base plus real-time web search results for more accurate renderings — useful for infographics and data visuals that need to be factually right.
- Upscales to 2K/4K in seconds: upscaling is built in, so there's no separate tool needed after generation.
What it’s good for
- Series of product or character shots that need the same face, product or details to stay consistent across images.
- Posters, marketing material and infographics that need accurate, legible text baked into the image — Google also positions it for translating and localizing that text.
- The same image reformatted for different platforms, by adjusting the aspect ratio to fit the placement.
Prompting tips
- Cover five elements: style (illustration/photo/watercolor…), subject (appearance, clothing, pose), setting (location + atmosphere), action (what's happening), and composition (shot type + angle) — the more detail you give, the closer the result gets to what you pictured.
- Use one of the five official editing moves: swap the character, adjust the composition (a new angle), change the action, swap the setting, or rethink the style — pick one at a time and describe exactly what should change.
- Put on-image text in quotes: quote the words you want rendered and pair them with a typography note, like "bold sans-serif" or "neon cursive."
- Name your characters and objects when using reference images: upload clear references and give people or objects consistent names in the prompt so they stay recognizable across generations.
- Ask directly for multiple variations: e.g. "three different product photography styles" or "four color palettes," to get several versions in one pass.
- State aspect ratio and intended use together: e.g. "vertical social post" or "widescreen backdrop," and use the 2K/4K upscale option when you need the extra resolution.
Illustration style. Subject: an orange cat named "Mochi" wearing a vintage dive mask and a yellow
raincoat. Setting: a seaside town at dawn, cobblestone streets, light fog. Action: Mochi is pushing
a small bicycle toward a lighthouse. Composition: medium shot, slightly high angle.
Hand-lettered text in the upper left reads "Good Morning." Vertical 9:16, for a social media poster.
Known limits
- Small faces, exact spelling and fine detail can still go wrong: Google states the model can struggle with small faces, accurate spelling, and fine details in images.
- Infographics can get misread: when generating infographics or data visualizations, it can sometimes misinterpret the underlying information.
- Advanced edits are less reliable: masked editing, major lighting changes (like day to night), or blending multiple images can sometimes produce unnatural results, per Google.
Using it in Nomi
On the generation canvas, add a image card, pick “Nano Banana 2” in the model picker and choose a mode (Text to image, Image editing, Image to image). Add your references, write the prompt and generate. In the storyboard you can pick it per shot. The first time, connect any one of the providers above under Settings → Models.
