Imagifly — Nano Banana 2

Nano Banana is Google's product name for Gemini's native image generation, not a separately published open model. Google's own mapping runs: Nano Banana is Gemini 2.5 Flash Image, Nano Banana Pro is Gemini 3 Pro Image, and Nano Banana 2 is Gemini 3.1 Flash Image, announced on 26 February 2026 and generally available through Gemini Enterprise Agent Platform on 28 May 2026.

  • Text to Image
  • Image to Image

From 7 credits per image

Cinematic city artwork sample for Nano Banana 2
Sample output: Cinematic city artwork sample for Nano Banana 2

About Nano Banana 2

Nano Banana is Google's product name for Gemini's native image generation, not a separately published open model. Google's own mapping runs: Nano Banana is Gemini 2.5 Flash Image, Nano Banana Pro is Gemini 3 Pro Image, and Nano Banana 2 is Gemini 3.1 Flash Image, announced on 26 February 2026 and generally available through Gemini Enterprise Agent Platform on 28 May 2026.

It generates and edits in the same conversation: multi-turn edits keep characters and scenes consistent, reference images are accepted, output comes at 1K, 2K or 4K (4096x4096 for 4K square), and prompts can ground results in Google Search or Image Search. Every output carries a SynthID watermark and C2PA content credentials.

On 6 October 2026 Google published Nano Banana 2.1 (gemini-nano-banana-2.1); the docs now recommend it for all new projects and describe Nano Banana 2 as the previous-generation high-efficiency model. The weights are not public, so the model is only reachable through the Gemini API or Vertex AI.

What it is good at

  • Text-to-image plus multi-turn conversational editing with reference images
  • 1K, 2K and 4K output; 4K square is 4096x4096
  • Text rendering is the headline capability for posters, infographics and menus
  • Google Search grounding, with SynthID and C2PA on every output

Technical specs

Developer
Google (Gemini API, Vertex AI)
Released
2026-02-26
Latest version
Nano Banana 2.1 (2026-10-06)
License
Proprietary, API-only
Architecture
Not disclosed by Google (Gemini 3.1 Flash Image)
Parameters
Not disclosed (proprietary, API only)
Task
Text-to-image and image editing (multi-turn)
Recommended settings
The docs allow up to 14 reference images (up to 10 objects plus 5 characters) and aspect ratios from 1:1 to 21:9, including extreme 1:4, 4:1, 1:8 and 8:1.

Source: Gemini API image generation guide