live

MODEL IDgoogle:nano-banana@2.1

Nano Banana 2.1

by October 6, 2026

Nano Banana 2.1 is Google's updated image generation and editing model in the Nano Banana 2 family. It improves graphic composition and subject consistency, and it follows complex prompts and edit instructions more accurately. It works from text alone or from as many as fourteen reference images plus one reference video, and it can ground a generation in live web and image search so the result reflects current facts and real visual references. It generates at 1K, 2K and 4K, with three levels of thinking that trade speed for reasoning depth, which suits layout-heavy design work, recurring characters and products, and precise multi-step edits.

Subject consistency

How to keep one product or character identical across a series with Nano Banana 2.1: one reference into new scenes, extra views, and illustrated characters.

Introduction

Generate a product shot you like, change the prompt to move it into a new scene, and the model hands you a different product. Every request starts from nothing, so the stitching or the exact shade of green drifts as soon as anything around it changes. For a catalog or a campaign, that drift makes the set unusable.

Nano Banana 2.1 takes the subject from an image instead of a description. Pass it in inputs.referenceImages, describe the new scene in positivePrompt, and the model carries the subject into that scene. One request accepts up to 14 reference images.

A sage green ribbed hard-shell suitcase with tan leather handles standing on a wet airport curb at dawn beside the open trunk of a yellow taxi, a traveler's hand resting on its top handle

Built from 1 reference

Reference
A studio packshot of a hard-shell carry-on suitcase in matte sage green with horizontal ribs, tan leather side and top handles, brass corner rivets, and four black spinner wheels, standing upright at a three-quarter angle on a plain white backdrop. Soft even studio lighting, subtle contact shadow, photoreal e-commerce photography, no logos, no text.

The reference is a packshot on white, and the prompt spends its words on the airport curb. Identity comes from the image, and the scene comes from the text.

This guide covers the request, holding a character across a campaign, adding references for the sides a scene will show, and keeping an illustrated character on model. Putting several different subjects into one frame is covered in multi-reference composition.

The request

Attach the subject as a reference, and the rest is an ordinary generation:

Try in Playground

import { createClient } from '@runware/sdk'

const client = await createClient({ apiKey: process.env.RUNWARE_API_KEY })
await client.connect()

const [result] = await client.run({
  model: 'google:nano-banana@2.1',
  positivePrompt: 'The suitcase from the reference image standing on the curb outside an airport terminal at dawn, next to the open trunk of a yellow taxi, with a traveler\'s hand resting on its top handle. Wet pavement reflecting the terminal lights, soft blue morning light with warm highlights. Keep the suitcase\'s sage green shell, horizontal ribs, tan leather handles, brass rivets, and black wheels identical. Photoreal travel campaign photography, shallow depth of field.',
  width: 2528,
  height: 1696,
  inputs: {
    referenceImages: [
      'https://im.runware.ai/image/os/a14d18/ws/2/ii/6b0d3e95-7a12-4c8f-9d46-b3e5f1a8c270.jpg'
    ]
  }
})

import asyncio
import os

from runware import Runware


async def main():
    async with Runware(api_key=os.environ["RUNWARE_API_KEY"]) as client:
        results = await client.run({
            "model": "google:nano-banana@2.1",
            "positivePrompt": "The suitcase from the reference image standing on the curb outside an airport terminal at dawn, next to the open trunk of a yellow taxi, with a traveler's hand resting on its top handle. Wet pavement reflecting the terminal lights, soft blue morning light with warm highlights. Keep the suitcase's sage green shell, horizontal ribs, tan leather handles, brass rivets, and black wheels identical. Photoreal travel campaign photography, shallow depth of field.",
            "width": 2528,
            "height": 1696,
            "inputs": {
                "referenceImages": [
                    "https://im.runware.ai/image/os/a14d18/ws/2/ii/6b0d3e95-7a12-4c8f-9d46-b3e5f1a8c270.jpg"
                ]
            }
        })


asyncio.run(main())

curl https://api.runware.ai/v1 \
  -H "Authorization: Bearer $RUNWARE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '[
    {
      "taskType": "imageInference",
      "taskUUID": "e1a4c7f2-5b93-4d08-8c6e-2f9d0a3b7c54",
      "model": "google:nano-banana@2.1",
      "positivePrompt": "The suitcase from the reference image standing on the curb outside an airport terminal at dawn, next to the open trunk of a yellow taxi, with a traveler's hand resting on its top handle. Wet pavement reflecting the terminal lights, soft blue morning light with warm highlights. Keep the suitcase's sage green shell, horizontal ribs, tan leather handles, brass rivets, and black wheels identical. Photoreal travel campaign photography, shallow depth of field.",
      "width": 2528,
      "height": 1696,
      "inputs": {
        "referenceImages": [
          "https://im.runware.ai/image/os/a14d18/ws/2/ii/6b0d3e95-7a12-4c8f-9d46-b3e5f1a8c270.jpg"
        ]
      }
    }
  ]'

runware run google:nano-banana@2.1 \
  positivePrompt="The suitcase from the reference image standing on the curb outside an airport terminal at dawn, next to the open trunk of a yellow taxi, with a traveler's hand resting on its top handle. Wet pavement reflecting the terminal lights, soft blue morning light with warm highlights. Keep the suitcase's sage green shell, horizontal ribs, tan leather handles, brass rivets, and black wheels identical. Photoreal travel campaign photography, shallow depth of field." \
  width=2528 \
  height=1696 \
  inputs.referenceImages.0=https://im.runware.ai/image/os/a14d18/ws/2/ii/6b0d3e95-7a12-4c8f-9d46-b3e5f1a8c270.jpg

{
  "taskType": "imageInference",
  "taskUUID": "e1a4c7f2-5b93-4d08-8c6e-2f9d0a3b7c54",
  "model": "google:nano-banana@2.1",
  "positivePrompt": "The suitcase from the reference image standing on the curb outside an airport terminal at dawn, next to the open trunk of a yellow taxi, with a traveler's hand resting on its top handle. Wet pavement reflecting the terminal lights, soft blue morning light with warm highlights. Keep the suitcase's sage green shell, horizontal ribs, tan leather handles, brass rivets, and black wheels identical. Photoreal travel campaign photography, shallow depth of field.",
  "width": 2528,
  "height": 1696,
  "inputs": {
    "referenceImages": [
      "https://im.runware.ai/image/os/a14d18/ws/2/ii/6b0d3e95-7a12-4c8f-9d46-b3e5f1a8c270.jpg"
    ]
  }
}

Response

[
  {
    "taskType": "imageInference",
    "taskUUID": "e1a4c7f2-5b93-4d08-8c6e-2f9d0a3b7c54",
    "imageUUID": "39f8a1c6-d204-4b7e-a85c-0e6d9b2f4a17",
    "imageURL": "https://im.runware.ai/image/os/a14d18/ws/2/ii/39f8a1c6-d204-4b7e-a85c-0e6d9b2f4a17.jpg"
  }
]

The prompt has two jobs. Most of it describes what is new: where the subject is and what is happening around it. One sentence near the end names the features to hold, which points the model at the details that make this suitcase this suitcase. width and height set the new frame, since a campaign scene rarely shares the shape of the packshot.

One character across a campaign

A person works the same way as a product. Start from a reference that shows the face clearly, in even light:

A studio portrait of a woman with a dark brown high ponytail and a small scar through one eyebrow, smiling with her arms crossed, wearing a teal racerback top and black leggings against a light gray backdrop — Reference

Each image below used that portrait as its only reference:

The same woman in her teal top leading a group fitness class in a bright studio, in a lunge with both arms raised, with four participants following behind her — Leading a class

The same woman running along a misty riverside path at sunrise in a white long-sleeve top and black shorts — A different outfit

A close-up of the same woman laughing with her head turned to the right and a white towel around her neck, in a warmly lit locker room — A close-up

A different medium

The woman from the reference image as a flat vector avatar illustration for an app profile: head and shoulders, thick clean outlines, flat color fills, teal top, centered on a solid pale yellow background. Keep her ponytail, her face shape, and the scar through her left eyebrow recognizable. No text.

Every prompt names one distinguishing mark, the scar through her eyebrow. A mark like that gives the model a detail to hold that survives a new outfit and even a new medium. Clothing is part of the reference too, so say so when it should change: the riverside prompt asks for a white top "instead of the teal top".

Covering the sides a scene will show

A reference only carries what it shows. When a scene turns the subject around, the model has to invent the hidden side. The fix is a second reference that shows it. Here are two views of one delivery van, where the rear doors carry a graphic that the side view cannot reveal:

Side reference

A studio photograph of a compact electric delivery van in mint green with a white roof, seen from the front left three-quarter angle. A wide white stripe runs along the side, with the wordmark "FERNWAY" in dark green capitals on the side panel. Plain light gray backdrop, soft even lighting, photoreal, no people.

The same mint green van seen from directly behind, its two rear doors carrying a large white fern leaf illustration above the words FRESH TO YOUR DOOR — Rear reference

The two street shots below share one prompt. The first request sent only the side reference, and the second sent both:

Side reference only

The mint green delivery van from the reference driving away from the camera down a tree-lined suburban street in late afternoon light, seen from directly behind with its rear doors clearly visible. Keep the van's paint, roof, and graphics consistent with the reference. Photoreal, slight motion blur on the road.

The same mint green delivery van driving away down a tree-lined suburban street, its rear doors showing the large white fern leaf and FRESH TO YOUR DOOR — Side and rear references

The prompt never mentions the fern. The rear graphic came from the second reference, and with the side view alone the model repeated the side wordmark on the doors. Build the reference set from the views your scenes will show: the front, plus any side that carries detail the front cannot reveal.

An illustrated character across pages

The same mechanism holds a drawn character, where the reference also carries the style, from the line to the palette.

A gouache character design of a small upright fox with large round eyes and a white-tipped tail, wearing a yellow raincoat with one red button and red rain boots, on a cream background — Character sheet

A gouache illustration of the fox in the yellow raincoat kneeling in a vegetable garden, planting a seedling with a trowel between two red watering cans — Page 1

A gouache illustration of the fox in the yellow raincoat riding a small blue bicycle through a puddle on a rainy village street with glowing shop windows and no signs — Page 2

A gouache illustration of the fox in the yellow raincoat sitting under a large oak tree reading a red book as autumn leaves fall — Page 3

Each page prompt opens with "in the exact style of the reference image". For a drawn character the style is half the identity, and naming it stops the model from redrawing the fox in a cleaner or more detailed look than the sheet.

Tips

Start from a clean reference. A plain background and even light, with the whole subject in frame, give the model the most to hold.
Spend the prompt on the scene. The reference carries the subject, so the prompt describes the place and the action.
Name the features that must hold. One closing sentence, such as "keep the sage green shell and the tan leather handles identical", points the model at the details that matter.
Give a person one distinguishing mark. A scar or a streak of gray hair survives a new outfit and a new medium.
Say when clothing should change. The outfit is part of the reference until the prompt replaces it.
Add a reference for every side a scene shows. One image covers one viewpoint.
Ask for the reference's style on illustrated work. "In the exact style of the reference image" keeps the line and the palette on model.