Editing-first iteration
Google positions the model for generation and editing together, so drafts, revisions, and restyles can stay in one workflow.
Google image model · Live workspace
Nano Banana 2.1 image generator
Generate and edit images with Google’s newest Nano Banana model, tuned for multimodal image generation and editing with strong cost performance.
Gallery images come from the Nano Banana 2.1 model page on OpenArt, recompressed and served from this site's CDN. Validate them with your own tests.
01
Create with Nano Banana 2.1
Nano Banana 2.1 is selected by default, and you can switch to another available image model from the model menu.
AI Image
Results
Creation historyGenerated content will appear here
02 · Overview
Verified model and platform facts
Facts reviewed:
- Provider
- Official model ID
- gemini-nano-banana-2.1
- Inputs
- Text and images
- Platform output sizes
- 1K, 2K, 4K
- Platform reference limit
- Up to 14 images
- Platform status
- Live
What is Nano Banana 2.1?
Nano Banana 2.1 is Google’s newest Nano Banana image model. Google’s model documentation lists the model ID gemini-nano-banana-2.1, marks it as generally available with a release date of 2026-10-06, and describes it as optimized for multimodal image generation and editing with strong cost performance.
Google documents image generation and editing (including multi-turn edits), interleaved images and text, image generation from video input, virtual try-on, and Content Credentials (C2PA) support, plus Google Search and Image Search grounding. The same documentation lists up to 14 images per prompt, resolutions of 1K, 2K, and 4K, and 15 aspect ratios including 9:21. It also states that seed, topK, logprobs, temperature, and topP are not supported and return an API error when set.
The workspace on this page is the platform’s own configuration: text-to-image and image-to-image are open, the aspect-ratio selector offers 14 fixed formats (the official 9:21 is not exposed here), the resolution selector offers 1K, 2K, and 4K with 1K as the default, and the current platform pricing is 3, 4, and 6 points per result by resolution. Platform controls come from the live task catalog, not from Google’s documentation.
03 · Capabilities
Why use Nano Banana 2.1
An editing-focused, cost-efficient choice for everyday generation, reference-based editing, and repeatable production drafts.
Structured instruction following
Layer subject, composition, lighting, style, required text, and must-keep constraints; Google documents thinking support for complex prompts.
Reference-based editing
Upload up to 14 reference images on this platform to restyle, recompose, or localize an existing visual.
Wide-format layouts
The platform exposes 14 fixed ratios, including 21:9 and extreme banner formats, for heroes, posters, and vertical crops.
1K to 4K output
Pick the platform resolution that matches drafting, publishing, or close inspection, with points shown before submission.
In-image text layouts
Draft posters, cards, and infographic-style graphics with rendered copy, then proofread every character before publishing.
Search grounding
Google documents Google Search and Image Search grounding for real-world context; access depends on the integration and task.
Provenance signals
Google lists Content Credentials (C2PA) support for generated images, alongside the platform’s own review guidance.
03A · Model comparison
Nano Banana 2.1 vs Nano Banana 2 vs Nano Banana 2 Lite
Platform facts from the live task catalog. Points are per result; your workspace estimate is authoritative at submission time.
| Nano Banana 2.1 | Nano Banana 2 | Nano Banana 2 Lite | |
|---|---|---|---|
| Points — text-to-image | 3 (1K) · 4 (2K) · 6 (4K) | 4 flat (+≈1 pt per reference image) | 2 (1K only) |
| Points — image-to-image | 3 (1K) · 4 (2K) · 6 (4K) | 3 (1K) · 4 (2K) · 6 (4K) | 2 (1K only) |
| Output resolutions | 1K / 2K / 4K | 1K / 2K / 4K | 1K only |
| Open modes here | Text-to-image, image-to-image | Text-to-image, image-to-image | Text-to-image, image-to-image |
| Positioning | Newest generation, official GA 2026-10-06 | Previous generation, still live | Entry tier for lightweight 1K drafts |
04 · Use cases
Task types to try
Suggested task coverage for this model. Gallery images come from the Nano Banana 2.1 model page on OpenArt; validate them with your own tests before publishing.
Gallery images come from the Nano Banana 2.1 model page on OpenArt, recompressed and served from this site's CDN. Validate them with your own tests.
OpenArt showcase01
Posters and social visuals
Campaign posters, channel banners, and feed graphics where rendered text and layout hierarchy matter.
OpenArt showcase02
Product scene images
Studio-style product scenes with controlled lighting and copy space; verify labels and geometry before use.
OpenArt showcase03
Reference-based edits
Swap backgrounds, restyle scenes, or recompose layouts while asking the model to keep specified elements fixed.
OpenArt showcase04
Text-heavy layouts
Menus, cards, and infographic drafts with in-image copy; proofread every word and check small-size legibility.
OpenArt showcase05
Character consistency scenes
Recurring people or mascots across multiple scenes using references; inspect identity details after each generation.
05 · Prompts
Nano Banana 2.1 prompt examples
Six starter briefs covering product scenes, text layouts, reference edits, character consistency, and wide banners. Prompts are editable; sample images come from the Nano Banana 2.1 model gallery on OpenArt, recompressed and hosted on this site's CDN.
Prompt 01
Iced tea product scene
An editable starter brief. Open the full prompt, adapt it to your task, and review the output before use.
A 500 ml iced citrus tea in a clear PET bottle standing on wet slate, fine condensation droplets, soft window light from the left, a blurred café background, the right third kept empty for a headline. Photorealistic product photography. No logos and no text.
Prompt 02
Poster with exact text
An editable starter brief. Open the full prompt, adapt it to your task, and review the output before use.
A minimal concert poster using exactly these three lines: "STATIC / FLOW", "SAT 18 OCT", "HALL 7". Make the first line the largest, the date second, and the venue small but readable. Black and amber palette, one abstract waveform, generous margins, and no other words.
Prompt 03
Reference-based background swap
An editable starter brief. Open the full prompt, adapt it to your task, and review the output before use.
Using the uploaded product photo, replace the background with a sunlit modern kitchen counter and add a soft morning shadow under the product. Keep the product shape, label, color, and viewing angle exactly unchanged.
Prompt 04
Consistent character triptych
An editable starter brief. Open the full prompt, adapt it to your task, and review the output before use.
Using the uploaded portrait, render the same character in three panels: reading in a library, walking a rainy street under a transparent umbrella, and cooking in a warm kitchen. Keep the face, hairstyle, and outfit identical across all three panels. Flat illustration style, muted palette.
Prompt 05
Ultrawide 21:9 banner
An editable starter brief. Open the full prompt, adapt it to your task, and review the output before use.
An ultrawide 21:9 website hero banner for a fictional electric bicycle: dusk city waterfront, the bicycle placed in the right quarter, riders silhouetted on a lakeside path, teal and orange color grade, a large calm sky area on the left reserved for a headline. No text.
Prompt 06
Four-step infographic card
An editable starter brief. Open the full prompt, adapt it to your task, and review the output before use.
A simple infographic card that explains cold brew coffee in four numbered steps, each with a small icon and a short caption: grind, steep 12 hours, filter, serve over ice. Cream background, one accent color, clear visual hierarchy, small but readable text.
06 · Workflow
How to use Nano Banana 2.1
- 01
Write a structured prompt
Separate subject, action, composition, lighting, style, required text, and exclusions. For edits, state what may change and what must stay fixed.
- 02
Add references when needed
For image-to-image, upload only images that clearly communicate what to preserve or change; the platform allows up to 14.
- 03
Set ratio and resolution
Choose one of the 14 fixed aspect ratios and 1K, 2K, or 4K output, then read the point estimate on the generate button.
- 04
Inspect before publishing
Check text, faces, hands, object counts, product geometry, and spatial relationships; regenerate or refine instead of shipping unchecked output.
Responsible use
Limits, failures, and recovery
Google’s documentation states that seed, topK, logprobs, temperature, and topP are not supported and return an API error when set, so this workspace does not expose them. If a task fails after copying settings from another tool, remove those parameters and retry.
Google’s model card lists known limitations: small text is often blurry at 1K, long paragraphs and dense pages can render poorly, character consistency between input images and output is not always perfect, masked or doodle-based edits may only partially follow instructions, the original pose can persist in rare editing cases, left and right can be confused, and occasional slowness or timeouts can occur. Recovery: shorten and structure the prompt, reduce reference count, switch resolution, or retry later.
User discussions have raised individual experiences about prompt following, portrait likeness compared with references, and whether output matches Nano Banana Pro in realism. These are individual reports, not verified conclusions; treat them as a checklist — verify instruction compliance, identity similarity, and realism against your own requirements before publishing. Rights review for uploads and outputs remains your responsibility.
Sources
Sources and review boundary
Model identity, capabilities, and limits were checked against Google’s current model documentation and model card on 2026-10-07; platform modes, controls, and point prices come from the live task catalog.
07 · FAQ
Nano Banana 2.1 FAQ
What is the official model ID of Nano Banana 2.1?
Google’s model documentation lists the ID gemini-nano-banana-2.1, marks the model as generally available, and gives a release date of 2026-10-06.
Which modes are open on this platform?
The live platform catalog opens text-to-image and image-to-image, both priced by resolution. Google also documents video-input generation and virtual try-on, but those are not open in this workspace.
How many points does one result cost?
The current platform reference price is 3 points per result at 1K, 4 points at 2K, and 6 points at 4K, for both open modes. The estimate shown in the workspace before submission is authoritative.
How many reference images can I upload?
This platform allows up to 14 reference images, matching Google’s documented per-prompt maximum of 14 images.
How do I choose between Nano Banana 2.1, Nano Banana 2, and Nano Banana 2 Lite?
At 1K, 2.1 costs 3 points per result versus 4 for Nano Banana 2 text-to-image and 2 for Lite, which only outputs 1K. For editing-heavy work and wide layouts, 2.1 is the newer generation with documented multi-turn editing support; for simple 1K drafts, Lite is the cheapest entry. See the comparison table above for the full platform picture.
Can Nano Banana 2.1 render readable text in images?
Google documents improved text rendering for posters and diagrams, while its model card notes that small text at 1K and long paragraphs can still be blurry or wrong. Always proofread rendered copy at the size it will be published.
What are common failure modes and how do I recover?
Unsupported sampling parameters (temperature, topP, topK, seed, logprobs) return API errors; remove them and retry. Overly long or layered prompts, reference overload, and rare upstream slowness can also fail or time out; simplify the prompt, reduce references, change resolution, or retry later.
Will faces stay consistent with my reference photos?
Google’s model card states that character consistency between input images and generated output is not always perfect. Some users report likeness drift in individual cases. Verify identity details after every generation before publishing.
Does output carry content credentials?
Google’s model documentation lists Content Credentials (C2PA) support for this model. Provenance signals do not replace usage-rights review for your uploads and outputs.
Is this page an official Google website?
No. This is an independent creation platform that integrates the model through the API gateway. It is not operated by or affiliated with Google.
Related models and tools
Nano Banana 2
Compare the previous Nano Banana generation, still live on this platform.
Nano Banana Pro
Use Google’s Pro image model for more complex professional asset work.
Reverse-engineer a prompt from any image before generating a new one.
Compare image models and choose a workflow from one tool page.
Check plans and credit packages that cover Nano Banana 2.1 results.
Create next
Try Nano Banana 2.1 Create with Nano Banana 2.1
Return to the workspace with Nano Banana 2.1 selected by default, then review every result before publishing.