跳转到主要内容
By Google · August 26, 2025

Nano Banana

Google's fast everyday image model — the one to reach for when you want a usable result in under five seconds.

Vendor
Google
Released
Aug 26, 2025
Max resolution
1K
Best for
Speed + iteration
01 / Overview

What is Nano Banana?

Nano Banana is the unofficial name (originated as a team nickname at Google DeepMind, later confirmed publicly) for Gemini 2.5 Flash Image, released August 26, 2025 and GA on Vertex AI in October. It is the default model on CVY.AI because it strikes the best balance of speed, multilingual quality and per-image cost for everyday work.

It is the default model for most generations unless you pick a heavier option — the Gemini 2.5 Flash Image model from Google, tuned for speed and low per-image cost.

Use it for fast iteration, character consistency work, multilingual prompts, and any case where the image is going to be scrolled past rather than printed. Reach for Nano Banana Pro or GPT Image 2 when the output has to survive close inspection.

02 / Capabilities

What Nano Banana is good at

  • Fast — 1-2 seconds typical

    Google's published latency for Gemini 2.5 Flash Image puts most generations in the 1-2 second range, with edge cases under 10. Faster than every other model on the platform, which is why it is the default for the homepage trial and the workhorse for batch iteration.

  • Character consistency across prompts

    The headline feature at launch. Supply a reference image and Nano Banana keeps the same character recognizable across new poses, scenes and outfits — useful for storyboarding, product variation shots, and "same person, different setting" social content.

  • Multi-image fusion

    Blend multiple reference images into one output. Useful for "this character + this setting + this lighting" composites that would normally take three rounds of prompt-editing in a one-shot model.

  • Multilingual prompt understanding

    Reads prompts in English, Chinese, Japanese and Korean without an explicit translation step. The Chinese results are noticeably better than what GPT-Image-1.5 used to produce — same family that powers Gemini app for non-Latin users.

  • Cheapest non-trivial option here

    At 2 credits per 1K image, it is the lowest-cost named model on cvy.ai with a Google-quality engine behind it. Margin is thin by design so we can keep the homepage trial free and the Pro tier credits stretch further.

03 / When to use

When to reach for Nano Banana

Use it for
  • +Iterating fast across many prompts — speed compounds when you generate dozens of variants.
  • +Social posts, blog heroes, product mockups, and other "draft quality is enough" work.
  • +Anything where keeping the same character recognizable across multiple frames matters (storyboards, character variations).
  • +Chinese, Japanese, or Korean prompts — multilingual quality is well above the older models.
Skip it for
  • In-image text — Nano Banana renders text approximately but not reliably. Use GPT Image 2 instead.
  • Hero / print / "this image will be enlarged and scrutinized" work — Nano Banana Pro is sharper at the same composition.
  • Long, multi-constraint prompts with strict layout requirements — reasoning matters there; pick GPT Image 2.
04 / Pricing

How much does Nano Banana cost?

ResolutionCredits
1K2
2K2
4K2

1 credit ≈ one 1K Nano Banana generation. See pricing for monthly credit bundles.

05 / FAQ
How much does Nano Banana cost?+
Generating with Nano Banana costs 2 credits per image. Free accounts include a monthly credit allowance to get started, and paid plans add more.
How is this different from Nano Banana Pro?+
Same Google family, two tiers. Pro (Gemini 3 Pro Image, released Nov 20, 2025) is roughly 2.5× the cost and noticeably sharper with better text rendering. Regular Nano Banana is faster, cheaper, and stronger on multilingual prompts. Use Pro for finals, regular for drafts.
Does Nano Banana support image-to-image editing?+
Yes — upload one or more reference images and the model will use them for composition, character consistency, or style guidance. Multi-image fusion (mixing several references into one output) is a launch feature, not a later add-on.
Why is it so cheap?+
Google's API price for Gemini 2.5 Flash Image is low, and we pass it through with a small margin. There is no quality lock-in — feel free to try other models for the same prompt and compare.
06 / Related models

Compare other AI image models

Use these pages to pick between speed, output quality, text rendering, and prompt-language fit.

View all models →
07 / Get started

Ready to try Nano Banana?

Open the generator with Nano Banana preselected, or browse community work for inspiration.