Nano Banana 2.1
Nano Banana 2.1 Text-to-Image is a high-performance image generation model built upon a brand-new architecture—a significant evolution from Nano Banana 2.0. The model possesses exceptional natural language semantic understanding, enabling it to precisely translate complex text prompts into ultra-high-definition images featuring cinematic composition, nuanced lighting and shadows, realistic textures, and high-precision text rendering. It supports native resolutions up to 4K and customizable aspect ratios, demonstrating remarkable stability when handling complex scene compositions and the arrangement of multiple elements.
Related Models
Nano Banana 2.1 Text-to-Image
Nano Banana 2.1 Text-to-Image is an advanced text-to-image synthesis model built on the upgraded Nano Banana 2.1 architecture. Engineered to convert complex natural language prompts directly into high-fidelity visual assets, it excels at composition execution, intricate lighting mechanics, and sharp typography rendering. Operating with support for dynamic aspect ratios and native resolutions up to 4K, it ensures precise photorealistic textures and consistent stylistic control across diverse artistic styles.
Nano Banana 2 Economy
Nano Banana 2 Economy offers an ultra-low-cost deployment channel for the Nano Banana 2 image generation model architecture. Engineered for scalable visual content workflows, it delivers sharp photorealistic textures, dynamic style adaptation, and precise prompt obedience across resolution tiers up to 4K at a flat cost per output. While retaining the core image synthesis capabilities of Nano Banana 2, this Economy channel operates with dynamic concurrency management and standard execution priority compared to the Official channel. It stands as an optimal choice for high-volume graphic design, social media content scaling, and automated design batching where unit cost efficiency is paramount.
Nano Banana 2 Lite
Nano Banana 2 Lite is an ultra-fast, lightweight image generation model engineered for high-volume visual workflows and real-time interactive applications. Operating on an optimized Flash-Lite vision architecture, it achieves sub-2-second rendering latencies while dramatically lowering compute overhead. Despite its blazing speed, Nano Banana 2 Lite maintains vibrant color fidelity, accurate prompt comprehension, and reliable in-image typography rendering. It is the definitive workhorse for high-frequency social media assets, rapid design prototyping, interactive consumer apps, and automated e-commerce batch processing where instant generation speed and unit cost optimization take priority.
Nano Banana
Nano Banana is an ultra-fast, high-performance image generation model engineered for high-efficiency visual workflows. Built on next-generation vision technology, it excels at translating text prompts into crisp, vivid artwork, social media graphics, e-commerce assets, and stylized illustrations within seconds. Balancing rapid generation speed with rich texture quality and accurate prompt comprehension, Nano Banana provides digital creators, designers, and marketers with a dependable, cost-effective engine for high-volume content production and instant creative experimentation.
Read Me
Nano Banana 2.1 API
Nano Banana 2.1 is a high-fidelity multimodal AI image generation and editing model developed by Google, built on the upstream Gemini Nano Banana 2.1 model that reached GA on 2026-10-06. It combines text-to-image generation and image-to-image editing in a single endpoint, delivering strong subject and character consistency, precise instruction following, natural language understanding, multi-reference image guidance, and both local and global editing at very fast generation speeds.
On the iCreat platform, Nano Banana 2.1 is served through the asynchronous endpoint https://api.icreat.ai/v1/task/submit/google/gemini-nano-banana-2-1, with results retrieved from https://api.icreat.ai/v1/task/result. The endpoint follows a two-step asynchronous workflow (submit a task, then poll for the result using the same task_id). Output resolutions are available in three tiers — 1K, 2K, and 4K — priced at $0.04, $0.06, and $0.135 per image respectively, plus a $0.02 surcharge per reference image.
Model Positioning
Nano Banana 2.1 is positioned as a high-fidelity, general-purpose image generation and editing model for production workflows that demand strong subject consistency and precise prompt adherence. Its main endpoint handles both text-to-image and image-to-image/edit tasks in one place, so teams can generate fresh images and refine existing ones without switching integrations.
The model is part of the Nano Banana 2.1 family on iCreat, which also includes a dedicated text-to-image endpoint (google/gemini-nano-banana-2-1/text-to-image) with no reference image field, and an Economy channel (google/gemini-3-pro-image-economy) offering lower-cost, less stable availability. The main endpoint described here is the recommended choice when reference-image editing is required.
Core Capabilities
Subject and Character Consistency
Nano Banana 2.1 excels at preserving the identity of subjects and characters across generations and edits, making it suitable for recurring characters, product shots, and multi-turn editing sessions where visual continuity matters.
Precise Instruction Following
The model interprets detailed, natural-language prompts accurately, translating complex compositional and stylistic requirements into the intended output with minimal drift.
Natural Language Understanding
Prompts are understood as free-form natural language, including multi-line text, allowing users to describe scenes, edits, and constraints conversationally rather than through rigid templates.
Multi-Reference Image Guidance
Up to 4 reference images can be supplied per request, each up to 20MB, enabling guided generation and editing based on publicly accessible image URLs.
Local and Global Editing
The model supports both localized edits (targeted changes to specific regions) and global edits (whole-image transformations), giving fine-grained control over the editing workflow.
Ultra-Fast Generation
Generation is optimized for speed, supporting rapid iteration in creative and production pipelines.
Pricing
| Resolution | Price (USD) | Billing Unit |
|---|---|---|
| 1K | $0.04 | per image |
| 2K | $0.06 | per image |
| 4K | $0.135 | per image |
Total cost = price for the selected resolution + number of reference images × $0.02.
Reference image surcharge: $0.02 per image. When reference images are used (image-to-image / edit mode), each reference image incurs an additional $0.02. Pure text-to-image requests (no reference images) are charged only the resolution price.
Prices listed here reflect iCreat AI's standard rates. Some models may be eligible for promotional discounts or special offers. Final charges are based on the real-time prices displayed in each model's Playground.
Application Scenarios
- Recurring character and brand asset generation with consistent identity
- Product photography and e-commerce visuals requiring subject fidelity
- Multi-turn image editing with reference-image guidance
- Marketing and social media creative at 1K, 2K, or 4K resolutions
- Infographics and text-heavy visuals needing accurate text rendering
- Panoramic and wide-format compositions (including 21:9)
- Rapid creative iteration powered by ultra-fast generation
Model Comparison
Comparison Table 1: Nano Banana 2.1 vs Nano Banana 2.1 Text-to-Image Endpoint
| Feature | Nano Banana 2.1 (Main Endpoint) | Nano Banana 2.1 Text-to-Image Endpoint |
|---|---|---|
| Endpoint | google/gemini-nano-banana-2-1 |
google/gemini-nano-banana-2-1/text-to-image |
| Mode | Text-to-image + image-to-image/edit | Pure text-to-image |
| Reference Images | Up to 4 | Not supported (no field) |
| Reference Image Surcharge | $0.02 per image | None |
| 1K Price | $0.04 per image | $0.04 per image |
| 2K Price | $0.06 per image | $0.06 per image |
| 4K Price | $0.135 per image | $0.135 per image |
Note: Choose the main endpoint when reference-image editing is needed (adds $0.02 per reference image). For pure text-to-image, both endpoints share the same resolution pricing, so either can be used.
Comparison Table 2: Nano Banana 2.1 vs Nano Banana 2.1 Economy
| Feature | Nano Banana 2.1 (Standard Channel) | Nano Banana 2.1 Economy |
|---|---|---|
| Channel | Standard | Economy |
| 1K Price | $0.04 per image | $0.04 per image |
| 2K Price | $0.06 per image | $0.04 per image |
| 4K Price | $0.135 per image | $0.04 per image |
| Pricing Model | Per-resolution tier | Flat $0.04 per image (1K/2K/4K) |
| Reference Images | Up to 4 | Up to 11 |
| Reference Image Surcharge | $0.02 per image | — |
Note: For 2K pure text-to-image, the Economy channel is cheaper ($0.04 vs $0.06); for 4K, the standard model is $0.135 versus $0.04 for Economy. Choose based on resolution tier and whether reference images are required.
Why Choose Nano Banana 2.1?
- Unified text-to-image and image-to-image/edit in a single main endpoint
- Strong subject and character consistency across generations and edits
- Precise instruction following and natural language understanding
- Up to 4 reference images per request for guided generation and editing
- Three resolution tiers (1K / 2K / 4K) with transparent per-image pricing
- Ten aspect ratios, including a 21:9 panoramic option
- Ultra-fast generation for rapid iteration
- Simple two-step asynchronous API with a flat request body
Specifications
| Field | Value |
|---|---|
| Model Name | Nano Banana 2.1 |
| Developer | |
| Model ID | google/gemini-nano-banana-2-1 |
| Model Type | High-fidelity multimodal AI image generation and editing model |
| Endpoint | https://api.icreat.ai/v1/task/submit/google/gemini-nano-banana-2-1 |
| Task Result Endpoint | https://api.icreat.ai/v1/task/result |
| Authentication | Authorization: Bearer ${ICREAT_API_KEY} |
| Input Modalities | Text (prompt), Image (reference image URLs) |
| Output Modalities | Image |
| Supported Resolutions | 1K / 2K / 4K |
| Aspect Ratios | 1:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 21:9 |
| Reference Image Limit | Up to 4 per request, each up to 20MB, publicly accessible URL |
| Reference Image Surcharge | $0.02 per image |
| Prompt | Required; natural-language text, supports multi-line |
| Billing Unit | Per image (by resolution tier) + $0.02 per reference image |
| API Mode | Asynchronous two-step (submit task → poll for result) |
Architecture
Nano Banana 2.1 is built on Google's Gemini Nano Banana 2.1 upstream model, a multimodal image generation and editing architecture. The upstream model supports a context window of 131,072 tokens with a maximum output of 32,768 tokens, enabling rich prompt understanding and multi-reference guidance.
On iCreat, the model is exposed as an asynchronous two-step API. A request is submitted to the submit endpoint with a flat top-level body containing prompt, image, image_size, and aspect_ratio. The response returns a task_id, which is then used to poll the task result endpoint. While processing, the result is an empty array []; on success, the result contains an Image-type resource with a url and download_url, along with costUSD.
Notes
- The two steps use the same task_id; while processing,
resultis[]and polling should continue. FAILEDis a terminal state — check the prompt, image URLs, image format, aspect ratio, and other request parameters.- The request body uses flat top-level fields (not wrapped in
input/parameters). promptsupports multi-line text.- Reference images: up to 4 per request, each up to 20MB, and must be publicly accessible URLs.
image_sizesupports three tiers: 1K / 2K / 4K.aspect_ratiosupports 10 values; by default it follows the aspect ratio of the first uploaded reference image.- Reference image surcharge is $0.02 per image; total cost = price for the selected resolution + number of reference images × $0.02.
costUSDis returned only on success.
Frequently Asked Questions
How do I call Nano Banana 2.1?
Nano Banana 2.1 uses a two-step asynchronous workflow. Submit a task to https://api.icreat.ai/v1/task/submit/google/gemini-nano-banana-2-1 with Bearer authentication, then poll https://api.icreat.ai/v1/task/result with the returned task_id. The request body uses flat top-level fields: prompt, image, image_size, and aspect_ratio.
How is the cost calculated?
Total cost = price for the selected resolution + number of reference images × $0.02. Resolution tiers are 1K at $0.04, 2K at $0.06, and 4K at $0.135 per image. For example, a 2K request with 2 reference images costs $0.06 + $0.04 = $0.10. The costUSD field is returned only on success.
How many reference images are supported?
Up to 4 reference images per request, each up to 20MB, and each must be a publicly accessible URL. When image is an empty array, the request is treated as pure text-to-image.
Which resolutions and aspect ratios are supported?
Three resolution tiers: 1K, 2K, and 4K. Ten aspect ratios: 1:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, and 21:9. By default, the aspect ratio follows the first uploaded reference image.
What upgrades does it bring over Nano Banana 2?
Per Google's official positioning: better visual detail and prompt adherence, more accurate text rendering and infographic layout, stronger subject consistency in multi-turn editing, and fixes for tiling artifacts in panoramic 2K/4K compositions. Nano Banana 2 has been marked as deprecated by Google.
What is the difference from the Nano Banana 2.1 text-to-image endpoint?
This main endpoint supports reference-image editing with a $0.02 surcharge per reference image, while the text-to-image endpoint is pure text-to-image with no reference image fee. Both endpoints share the same 1K/2K/4K resolution pricing.