Nano Banana API pricing: what an image costs and how to pay less

Nano Banana is Google's official short name for Gemini 2.5 Flash Image, a fast text-to-image and image-editing model with strong character consistency, multi-image fusion, and natural-language editing. It went generally available in October 2025 and is now deprecated: for new work, use Nano Banana 2.1, the cheapest current Nano Banana, or Nano Banana 2 and Nano Banana Pro.

TL;DR: Nano Banana is Gemini 2.5 Flash Image, and it is deprecated. It accepted a prompt plus up to 20 reference images (10 MB max each) and ten aspect ratios from 1:1 through 21:9. Its successors keep that surface and add resolution choice: Nano Banana 2.1 from $0.02 per image, Nano Banana 2 (Gemini 3.1 Flash Image) from $0.034, and Nano Banana Pro (Gemini 3 Pro Image) from $0.07.

What is Nano Banana?

Nano Banana is Google's nickname (and now official short name) for the Gemini 2.5 Flash Image model — which is why some searches call it "Nano Banana 2.5"; that's this model, not the separate Nano Banana 2. It handles image generation and editing in one route: new images from a prompt, natural-language edits of existing images, and blends of multiple reference images into a single composition. It became generally available in October 2025 through the Gemini API, Google AI Studio, and Vertex AI.

The model is known for four capabilities Google highlighted at launch:

• Character consistency. Keeps the same person or object recognisable across multiple prompts and edits.
• Multi-image fusion. Blends up to 14 reference images into a single output (Pro) or up to 20 references on UnificAlly's route.
• Prompt-based editing. Change specific parts of an image using natural-language instructions.
• Visual template adherence. Follows a fixed layout (employee badges, real-estate cards, product mockups) across many generations.

Every Nano Banana output carries Google's invisible SynthID watermark for AI-content provenance.

What Nano Banana solved

Nano Banana fixed two things at once that had kept earlier text-to-image models out of production work:

1. Character identity drift. Before Nano Banana, asking a model to render the same character in three different scenes typically returned three different faces. Gemini 2.5 Flash Image holds identity stably across prompts.
2. Conversational editing. "Change the background to a sunset beach" or "give the person a winter coat" works as a one-shot edit, not a long prompt-engineering session.

That combination opened up product workflows that previously required expensive retouchers: catalogue stills with one consistent model across many scenes, branded character series, and template-driven assets like agent badges.

What you can do with Nano Banana

Inputs

• Prompt. Required text describing the image.
• Reference images. Up to 20 optional URLs, each up to 10 MB. Used for subject consistency, style transfer, or multi-image fusion.
• Aspect ratio. Choose from ten ratios spanning 1:1 through 21:9.

Output

• A single image at the chosen aspect ratio.
• SynthID invisible watermark embedded for AI provenance.
• Returned via task-based async polling.

Variants on UnificAlly

The original Nano Banana is deprecated. These are the current Nano Banana models; pick by what you need.

Use Nano Banana 2.1 for the cheapest drafts and most everyday work. Use Nano Banana 2 when an existing pipeline is tuned to it. Use Nano Banana Pro when you want Gemini 3 Pro's stronger text rendering and reasoning.

Nano Banana pricing and how it compares

Our current rates for every live Nano Banana model:

Each is billed per image at the resolution you request, with no subscription. The pricing page lists every model.

How to call Nano Banana

The API is async: POST a generation, then poll until the image URL is ready.

const API = 'https://api.unifically.com';
const headers = {
  Authorization: `Bearer ${process.env.UNIFICALLY_API_KEY}`,
  'Content-Type': 'application/json',
};

const start = await fetch(`${API}/v1/tasks`, {
  method: 'POST',
  headers,
  body: JSON.stringify({
    model: 'google/nano-banana-2.1',
    input: {
      prompt: 'A studio portrait of the same character from the reference image, now wearing a wool coat in a snowy alley',
      aspect_ratio: '3:4',
      resolution: '1K',
      image_urls: ['https://example.com/character-reference.jpg'],
    },
  }),
}).then((r) => r.json());

while (true) {
  await new Promise((r) => setTimeout(r, 2000));
  const task = await fetch(`${API}/v1/tasks/${start.data.task_id}`, { headers }).then((r) => r.json());
  if (task.data.status === 'completed') {
    console.log(task.data.output.image_url);
    break;
  }
  if (task.data.status === 'failed') throw new Error(task.data.error_message);
}

The original Nano Banana is deprecated, so the sample runs on google/nano-banana-2.1. Every Nano Banana model takes the same request: change model to google/nano-banana-2 or google/nano-banana-pro and keep the rest of the client code.

Things to know about Nano Banana

• Stuffing 20 references when you only need 2. Reference images compete for influence. Over-loading dilutes the effect of the most important one. Send the minimum that anchors the look.
• Asking for absurd aspect ratios for editing. Nano Banana works best at the ten supported ratios (1:1 through 21:9). Forcing exotic dimensions through cropping after the fact gives better results than fighting the model.
• Treating SynthID as optional. The invisible watermark is always embedded. If your downstream pipeline strips metadata, you lose the AI provenance signal, so keep it for compliance.
• Defaulting to Pro for everything. Pro doubles the per-image cost. Start on Nano Banana 2.1 at 1K for drafts. Promote to 4K or Pro only when the result survives review.
• Using Nano Banana for legible long-form text rendering. Short text usually works. Long paragraphs of in-image type are still better handled by Nano Banana Pro (Gemini 3 Pro Image), which Google specifically tuned for text rendering.

Nano Banana FAQ

What is Nano Banana?

Nano Banana is Google's Gemini 2.5 Flash Image model. It generates and edits images from natural-language prompts, holds character identity across edits, fuses up to 20 reference images, and includes a SynthID invisible watermark. It went generally available in October 2025.

How much does Nano Banana cost?

The original Nano Banana is deprecated. Its successors are priced by resolution: Nano Banana 2.1 is $0.02 at 1K, and Nano Banana 2 is $0.034 at 1K, $0.051 at 2K, $0.076 at 4K. Nano Banana Pro (Gemini 3 Pro Image) is $0.07 per image. Check the pricing page for current rates.

What aspect ratios does Nano Banana support?

Ten aspect ratios from 1:1 through 21:9, including portrait, landscape, square, and ultrawide options. The aspect ratio is set per generation request alongside the prompt and any reference images.

Can Nano Banana keep the same character across generations?

Yes. Character consistency is one of Nano Banana's core capabilities. Pass a reference image of the subject, then prompt new scenes; the model preserves facial features, body proportions, and distinctive traits across separate runs.

When should I use Nano Banana 2 or Nano Banana Pro instead?

Use Nano Banana 2 when you want explicit 1K, 2K, or 4K resolution control with resolution-priced variants. Use Nano Banana Pro when you need Gemini 3 Pro's stronger text rendering and multi-turn reasoning for studio-grade work.

Related reading on Nano Banana

• Nano Banana 2.1 API: the current, cheapest Nano Banana.
• Nano Banana model page: the deprecated original.
• Nano Banana 2: resolution-priced Gemini image generation.
• Nano Banana Pro: Gemini 3 Pro Image at 1K, 2K or 4K.
• GPT Image 2 and Flux.2: other image APIs on UnificAlly.