Google Nano Banana 2 Text-to-Image

Introduction

Nano Banana 2 Text-to-Image is Google’s prompt-only endpoint for high-frequency visual production. It is designed for creating marketing concepts, campaign key art, content covers, and multi-channel images directly from written requirements. Because the page does not include reference generation or image editing, it is best suited to users who can describe the intended result clearly in text.

On iCreat, users can enter prompts of up to 8,192 characters, choose from ten common aspect ratios, and generate at 1K, 2K, or 4K resolution. The endpoint uses fixed resolution-based pricing and is available through the Playground or the google/gemini-3-1-flash-image/text-to-image API.

Key Features

Prompt-only image generation

This endpoint accepts written prompts without requiring users to upload or organize reference images. Creative goals, subject relationships, composition, exact text, and publishing requirements can be combined in one brief, making it practical for rapid concept development, batch proposals, and visual production that does not depend on existing source assets.

Build complete compositions from written briefs

Nano Banana 2 can combine subjects, environments, camera direction, lighting, style, exact copy, and intended use inside one generated composition. When the prompt defines clear priorities, the model has more context for arranging subject placement, text areas, negative space, and supporting elements around the primary communication goal.

Plan layouts around the aspect ratio

The endpoint provides ten common aspect ratios, allowing users to choose the final placement before generation. Separate instructions can be written for vertical ads, square posts, landscape banners, or ultrawide assets, reducing the composition loss that often occurs when one finished image is forced into several different crops afterward.

Generate from 1K to 4K

Users can select 1K, 2K, or 4K output for early concepts, web publishing, presentations, or final delivery. The same prompt and ratio controls apply across the three tiers, so teams can validate a direction at a smaller size before generating a larger final version for production.

Fixed pricing for text-to-image work

The text-to-image endpoint is priced only by resolution: $0.07 for 1K, $0.105 for 2K, and $0.14 for 4K, with no additional quality tier. For teams that only need prompt-based generation, this structure is easier to budget than a complete endpoint that also includes reference inputs and editing workflows.

How to Use

1. Enter the prompt

This endpoint is dedicated to text-to-image generation. Enter a written brief to create a new image from scratch without uploading references or selecting an editing mode.

2. Define the composition

Describe the subject, environment, camera, lighting, style, exact copy, and intended use, then identify the most important content and any space that must remain open.

3. Select ratio and resolution

Choose the aspect ratio for the final placement, then select 1K, 2K, or 4K according to drafting, web publishing, presentation, cropping, or delivery requirements.

4. Submit the generation task

Generate directly in the Playground or call google/gemini-3-1-flash-image/text-to-image, poll the task status, and retrieve the completed image when processing succeeds.

Pricing

Nano Banana 2 Text-to-Image uses fixed per-image pricing based on output resolution, with no additional quality tier.

The 4K tier costs only $0.035 more than 2K, making it practical after the composition is approved and the final asset needs more cropping room or a larger delivery size.

Model Comparison

Nano Banana 2 Text-to-Image vs Gemini 3 Pro Image Text-to-Image

Nano Banana 2 targets high-frequency text-to-image production, while Gemini 3 Pro Image is positioned for complex professional visuals and finer creative control.

Nano Banana 2 Text-to-Image vs Seedream 5.0 Lite

Nano Banana 2 adds more aspect ratios and a 1K tier, while Seedream 5.0 Lite provides lower 2K and 4K pricing plus reference-based editing.

Parameters

Use Cases

Written-brief concept exploration

Turn campaign ideas, visual directions, or content requirements into several early images for internal discussion, direction selection, and proposal preparation without collecting reference assets first.

Headline-led campaign key art

Start from a headline, central subject, and placement requirements to create launch graphics, event covers, and promotional key art during the early creative stage.

Fictional product concepts

Describe a product’s function, materials, colors, and environment to create concept visuals for positioning discussions and early exploration before a real product asset library exists.

Presentation and editorial opening visuals

Generate presentation covers, feature headers, report openers, and editorial concepts with deliberate title areas, information space, visual focus, and an appropriate narrative mood.

Information graphics from written requirements

Describe steps, categories, labels, and hierarchy in the prompt to create simple infographic drafts, process visuals, and annotated explainers for concept validation.

Multi-channel composition planning

Generate separate square, vertical, landscape, and ultrawide compositions to test how the subject, headline, and negative space adapt across different publishing placements.

Prompt Tips

Define the visual hierarchy first. State the priority of the headline, primary subject, supporting elements, and negative space before adding style details.

Write exact image text. Put required copy in quotation marks and specify the language, position, size, alignment, and surrounding space.

Describe composition for the ratio. Select the final aspect ratio first, then define subject placement, text areas, and edge spacing instead of relying on later cropping.

State the image’s purpose. Tell the model whether the result is campaign key art, a presentation cover, an advertisement, or social content to guide information density and layout.

Notes

• aspect_ratio is required in the current schema, but its description mentions inheriting the first uploaded image’s ratio; this text-to-image endpoint has no image input, so always pass the ratio explicitly.
• The schema and Playground use 1K, 2K, and 4K, while the request example uses lowercase 1k; confirm whether the iCreat endpoint normalizes casing before production integration.
• Google’s native 0.5K option, extreme aspect ratios, Image Search grounding, and Thinking controls are not exposed in the current iCreat parameters and should not be presented as available endpoint features.
• The current endpoint has no visible watermark setting, but images generated by the Nano Banana family include an invisible SynthID watermark.

Related Models

Gemini 3 Pro Image Text-to-Image

Choose it for more demanding professional visuals, finer creative control, and higher-complexity text-to-image work.

Nano Banana 2

Choose it when you need reference-based generation, image editing, and up to 11 reference images in the complete Nano Banana 2 workflow.

Seedream 5.0 Lite

Choose it when lower unit pricing, equal 2K and 4K costs, reference inputs, and output-format control matter more.