comfyui-nano-banana

ComfyUI custom node that calls Nano Banana Pro (gemini-3-pro-image-preview) through Google AI Studio REST and returns an IMAGE output.

Inputs

• api_key (STRING, required): Google AI Studio API key.
• prompt (STRING, required): Prompt to generate image(s). Use ** to split into multiple prompts in one field (for example: prompt A ** prompt B ** prompt C).
• size (DROPDOWN, required): Output aspect ratio with 1K pixel dimensions. Default: 9:16 (768x1344). Options: 1:1 (1024x1024), 2:3 (832x1248), 3:2 (1248x832), 3:4 (864x1184), 4:3 (1184x864), 4:5 (896x1152), 5:4 (1152x896), 9:16 (768x1344), 16:9 (1344x768), 21:9 (1536x672).
• resolution (DROPDOWN, required): Output image size. Options: 1K, 2K, 4K.
• image (IMAGE, optional): Input conditioning image sent to the model.

Output

• image (IMAGE): Generated image tensor. If multiple prompts are provided via **, output is a batch with one image per prompt.

Install

1. Place this folder at custom_nodes/comfyui-nano-banana.
2. Install dependencies:

cd custom_nodes/comfyui-nano-banana
pip install -r requirements.txt

3. Restart ComfyUI.

Notes

• This node uses direct API key access (no Comfy API proxy).
• If the API returns multiple images, this node returns the first one.
• The node caches outputs in memory by (image_size, prompt, size, resolution) and skips API calls on cache hits.
• Prompt splitting uses ** as delimiter; empty segments are ignored.