comfyui-nano-banana-custom-node

ComfyUI Nano Banana

Gemini image generation for ComfyUI, built on Google's official Python SDK: google-genai.

中文文档

Overview

This custom node adds a Gemini Nano Banana Image node to ComfyUI.

It uses the official Gemini SDK flow instead of handwritten REST requests:

• from google import genai
• from google.genai import types
• client.models.generate_content(...)

Features

• Supports gemini-2.5-flash-image
• Supports gemini-3.1-flash-image-preview
• Supports gemini-3-pro-image-preview
• Supports text-to-image and reference-image workflows
• Supports response mode, aspect ratio, resolution, candidate count, seed, and people-generation control
• Supports per-node UI labels in English and Chinese
• Supports optional proxy control, timeout settings, and Google Search tool enablement
• Supports overriding the preset model with custom_model

Installation

Clone or copy this repository into your ComfyUI custom nodes directory:

ComfyUI/custom_nodes/ComfyUI-NanoBanana

Install the Gemini SDK in the same Python environment used by ComfyUI:

pip install -q -U google-genai

If your ComfyUI installation uses its own virtual environment, install the dependency there.

Restart ComfyUI after installation. Restarting is also required after frontend changes so the language-switch extension can load.

Node

• Node name: Gemini Nano Banana Image
• Category: Google AI/Gemini

Supported Models

• gemini-2.5-flash-image
• gemini-3.1-flash-image-preview
• gemini-3-pro-image-preview

If Google releases newer model names, you can enter them manually through custom_model.

Inputs

Core Inputs

• prompt: Prompt text for image generation or editing
• language: UI label language, english or chinese
• model: Preset Gemini image model
• custom_model: Overrides model when provided
• api_key: Gemini API key entered directly in the node
• api_key_env: Environment variable name used to read the API key

Generation Inputs

• response_mode: MODEL_DEFAULT, TEXT_AND_IMAGE, or IMAGE_ONLY
• aspect_ratio: Output aspect ratio, or default
• image_size: Output resolution, default, 1K, 2K, or 4K
• candidate_count: Number of image candidates to request
• seed: Random seed, use -1 to leave it unspecified
• person_generation: default, DONT_ALLOW, ALLOW_ADULT, or ALLOW_ALL

Network Inputs

• enable_google_search: Enables the Gemini Google Search tool
• proxy_mode: environment, direct, or manual
• proxy_url: Proxy URL used when proxy_mode is manual
• timeout_seconds: Request timeout for the SDK client

Reference Image Inputs

The node exposes up to 5 optional reference image inputs by default:

• image_1
• image_2
• image_3
• image_4
• image_5

Limits

• The node exposes 5 reference image ports by default
• gemini-2.5-flash-image is limited to 3 reference images on this node
• Other supported models currently use the node default limit of 5
• If Gemini returns images with mixed sizes, only the first size group is kept and a note is appended to the text output

API Key Resolution Order

The node resolves the API key in this order:

1. api_key
2. api_key_env
3. GEMINI_API_KEY
4. GOOGLE_API_KEY

Usage

Text to Image

Connect no reference images and provide only prompt.

Image to Image / Editing

Connect one or more reference images and describe the intended change in prompt.

Multi-Image Composition

Connect multiple reference images and describe how Gemini should merge or preserve them.

Notes

• Model availability and image-generation capabilities can change over time; follow the official Gemini documentation first
• The SDK may honor environment proxy settings unless you explicitly disable them with proxy_mode

References

• Gemini image generation docs: https://ai.google.dev/gemini-api/docs/image-generation
• Gemini quickstart: https://ai.google.dev/gemini-api/docs/quickstart
• Python SDK reference: https://googleapis.github.io/python-genai/