NanoBanana API Config
Image-gen settings for Gemini's image model
- llm_config
"NanoBanana" is Google's codename for its Gemini image-generation model - the one that got a lot of attention for how well it edits and composes images from a plain description. This node is the settings box for it. It's the image-generation counterpart to the generic LLM API Config: where that one configures a text model, this one configures NanoBanana with the knobs an image model needs - aspect ratio, resolution, and what kind of output you want back. The pack's README describes it plainly as a preset for Gemini 2.5 Flash Image.
Where it sits
Same three-node pattern as the rest of the art-venture LLM nodes. This Config wires into a Chat node, and you pair it with the Gemini API node that carries your Google API key. Config describes the image you want; the API node provides access; the LLM Chat node runs it and returns the picture. Because Gemini is a multimodal model, the chat can take your text (and optionally reference images) and hand back a generated image on the images output.
The inputs that matter
- model -
nano-banana,nano-banana-2, ornano-banana-pro. The newer/pro entries are the more capable revisions; pick based on what your account can reach. - aspect_ratio -
auto, or a fixed ratio from1:1through21:9. This is how you get portrait, landscape, or widescreen output without cropping later.autolets the model decide. - resolution -
1K,2K, or4K. Higher is sharper and slower (and, on a paid API, costs more per image). - modalities -
image+text,image, ortext. This controls what comes back: pickimage+textwhen you want the picture plus the model's commentary,imagefor just the render. - temperature - 0 to 1 (default 0.5), the usual creativity dial.
The output is an LLM_CONFIG for the Chat node.
Installing it
ComfyUI Manager → search comfyui-art-venture → install → restart. Or:
cd ComfyUI/custom_nodes
git clone https://github.com/sipherxyz/comfyui-art-venture
then restart ComfyUI. It's pre-installed on comfy.icu.
Common issues
The pack-wide one: a red "missing" node means art-venture isn't installed. Manager → "Install Missing Custom Nodes."
The real catch is that this is a cloud, closed-model node. Unlike a local checkpoint, NanoBanana runs on Google's servers, so you need a working Gemini API key, network access, and whatever billing/quota Google requires for the model tier you selected. No key or no access to that model, and the Chat node errors - the Config itself never complains because it's just settings. If your nano-banana-pro request fails, check whether your API key actually has access to that specific model before blaming the workflow. And if you set modalities: text and wonder where your image went - that's you telling it not to return one. For pictures, keep it on image or image+text.
Inputs (5)
| Name | Type | Default | Description |
|---|---|---|---|
| model | COMBO | nano-banana | 3 options: nano-banana, nano-banana-2, nano-banana-pro |
| temperature | FLOAT | 0.5000–1 | — |
| modalities | COMBO | image+text | 3 options: image+text, image, text |
| aspect_ratio | COMBO | auto | 11 options: auto, 1:1, 2:3, 3:2, 3:4, 4:3, +5 |
| resolution | COMBO | 1K | 3 options: 1K, 2K, 4K |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| llm_config | LLM_CONFIG | — |