GOOGLE · NANO BANANA 2.1
Nano Banana 2.1 is Google’s October 2026 update to Nano Banana 2. It renders text more accurately, keeps characters consistent across edits and handles extra-wide panoramas. Add up to 14 reference images and export at 1K, 2K or 4K.
Review results, manage assets and access generation history in AI Studio.
Every example below was actually generated with Nano Banana 2.1 on ImgAI, with the original prompt and settings. Click "Use this example" to load it into the creator.
Nano Banana 2.1 · 2K · 16:9
All eight diameters match published values, without search. Strictly, they are mean diameters — the prompt asked for equatorial ones. 2K.
A clean educational infographic titled "The Planets by Size": the eight planets of the solar system drawn side by side in order from the Sun, each labeled with its name and equatorial diameter in kilometers. Dark navy background, simple flat style, legible labels.
Nano Banana 2.1 · 2K · 4:1
An ultra-wide 2K panorama of a harbor town at golden hour, with no visible seams.
Ultra-wide panoramic photo of a Mediterranean harbor town at golden hour: pastel houses stacked on a hillside, small fishing boats in the bay, a church bell tower, calm water reflections.
Nano Banana 2.1 · 1K · 4:3
All three lines of lettering, opening hours included, came out exactly as written.
A hand-painted wooden shop sign above a bakery door. Main text "Golden Crumb", below it "Sourdough · Pastries · Coffee" and a small line "Mon–Sat 7:00–18:00". Gold and cream lettering on dark green, morning street light.
Nano Banana 2.1 · 1K · 3:4
Shop name and "豚骨ラーメン ¥980" are exact; the small decorative seals contain made-up characters.
日本拉面店的竖版海报：店名「一番麺屋」写在顶部，中间是一碗热气腾腾的豚骨拉面，底部写「豚骨ラーメン ¥980」。红黑配色，日式木版画风格，排版清晰。
Nano Banana 2.1 · 1K · 16:9
Both characters stay the same across three panels; in the last one the scarf moves to the snowman, as asked.
A three-panel comic about two friends: a tall girl with a red beanie and a short boy with round glasses and a yellow scarf. Panel 1: they roll a big snowball. Panel 2: they stack the snowman together. Panel 3: the snowman wears the boy's yellow scarf and both grin. Keep both characters identical in every panel. Clean line art, soft colors.
Nano Banana 2.1 · 1K · 16:9
Input: the mint robot from our Nano Banana 2 page. Recolored and moved to a beach; face, name badge and parcel kept.
Turn this robot sunflower yellow and place it on a sandy beach at sunset with gentle waves. Keep its face, shape, name badge and the parcel exactly the same.
Nano Banana 2.1 is an update to Gemini 3.1 Flash Image (Nano Banana 2). Google describes it as its latest high-efficiency model for image generation and conversational editing: Flash-level speed and cost, with clear gains in visual quality, prompt adherence, multi-turn character consistency and text rendering. It keeps up to four characters and ten objects consistent, and fixes the tiling artifacts Nano Banana 2 showed in panoramic ratios at 2K and 4K.
In Google’s own human-preference evaluation on its model card, Nano Banana 2.1 scored 1050 overall, ahead of Nano Banana 2 (990) and Nano Banana Pro (935). Google also lists its limits: small text is often blurry at 1K, and long paragraphs remain hard. On ImgAI it accepts up to 14 reference images, outputs 1K, 2K or 4K in 14 aspect ratios, and costs 22–58 credits per image.
Based on the model's published capabilities and what we actually generated, to help you judge whether it fits your task.
Google reports a clear jump in infographic quality and factuality over Nano Banana 2. Check every number before you publish.
Keep up to four characters recognizable across scenes and rounds of edits — mascots, comics, storyboards.
Short headlines and labels render cleanly, including Japanese and Chinese text.
4:1, 1:4, 8:1 and 1:8 without the seams Nano Banana 2 could show at 2K and 4K.
Specs come from the configuration currently available on this site; the creator always shows the live price. Failed generations are refunded automatically.
| Input | Text + up to 14 reference images |
| Resolution | 1K · 2K · 4K |
| Prompt length limit | 10,000 characters |
Supported aspect ratios
| Tier | 1K | 2K | 4K |
|---|---|---|---|
| Nano Banana 2.1 | 22 | 36 | 58 |
These come from real generations — read them alongside the original prompts in the examples.
Google notes small text can blur at 1K. If your image has captions or labels, render at 2K or 4K.
Put each title or label in quotes. Long paragraphs are still a weak spot for the model.
Give each character a fixed look ("red beanie, round glasses") and say "keep both characters identical" for multi-panel work.
Upload the earlier image and change one thing per request; the model is built for multi-turn editing.
Nano Banana 2.1 is Google’s October 2026 update to Gemini 3.1 Flash Image (Nano Banana 2) for image generation and conversational editing. On ImgAI it supports up to 14 reference images and 1K, 2K or 4K output.
Google lists better visual quality, prompt adherence, text rendering and multi-turn character consistency, plus fixed tiling artifacts in panoramic ratios at 2K and 4K. On ImgAI, Nano Banana 2 still offers the Google Search option; Nano Banana 2.1 does not yet.
22–58 credits per image on ImgAI, depending on resolution. At 1K it costs less than Nano Banana 2. Failed generations are refunded automatically.
In Google’s own preference evaluation it scored higher overall (1050 vs 935). Pro can still be the safer pick for very dense, multi-part layouts — try both on the same prompt.
Google’s model card notes blurry small text at 1K, difficulty with long paragraphs, imperfect consistency between input and output images, and occasional spatial confusion.
Google states that images generated by its models carry an invisible SynthID watermark. ImgAI adds no visible watermark.
Sign up and generate with credits. Results are saved to your library, ready to download or keep editing.