SorenWeile
7e3c8fb482
Introduce nbp_gemini_composite.py: a new ComfyUI API node that implements structured multi-reference Gemini image generation (product angles, character/outfit, additional references) with configurable model, seed, aspect ratio, resolution and advanced sampling params (temperature, top_p, top_k) and optional system prompt. Update __init__.py to import and merge NODE_CLASS_MAPPINGS and NODE_DISPLAY_NAME_MAPPINGS from the new module so the node is exposed. Update README to document the new advanced sampling options and mark system_prompt as advanced.
2.7 KiB
2.7 KiB
nbpOpenNode-ComfyUI
A custom ComfyUI node for generating images with Google Gemini image models, authenticated via your ComfyUI Cloud API key — no separate Google API key required.
Node: NBP Gemini Image 2
Generates (and optionally edits) images using Google's Gemini image models through the ComfyUI Vertex AI proxy. Auth is handled automatically by your ComfyUI account credentials.
Inputs
| Input | Type | Required | Description |
|---|---|---|---|
prompt |
String | Yes | Describe the image to generate or edits to apply |
model |
Dropdown | Yes | Gemini model to use (see models below) |
seed |
Int | Yes | Fixed seed for reproducibility (best-effort) |
aspect_ratio |
Dropdown | Yes | Output aspect ratio; auto matches the input image |
resolution |
Dropdown | Yes | 1K, 2K, or 4K — 2K/4K use Gemini's native upscaler |
temperature |
Float | Yes | Controls randomness (0.0–2.0). Lower = more predictable, higher = more creative |
top_p |
Float | Yes (advanced) | Nucleus sampling threshold (0.0–1.0). Lower = more focused output |
top_k |
Int | Yes (advanced) | Limits the token pool at each step (1–100). Lower = more deterministic |
response_modalities |
Dropdown | Yes (advanced) | IMAGE for image only; IMAGE+TEXT also returns model reasoning |
images |
IMAGE | No | Optional reference image(s) for editing or style guidance (up to 14, use Batch Images node for multiple) |
system_prompt |
String | No (advanced) | System-level instructions shaping the model's behavior |
Outputs
| Output | Type | Description |
|---|---|---|
IMAGE |
IMAGE | Generated image batch |
STRING |
STRING | Any text returned alongside the image (empty if IMAGE modality) |
Supported models
| Model | Notes |
|---|---|
gemini-3-pro-image-preview |
Highest quality, higher cost |
gemini-3.1-flash-image-preview |
Fast, lower cost (Nano Banana 2) |
gemini-2.5-flash-image-preview |
Previous generation preview |
gemini-2.5-flash-image |
Previous generation stable |
Installation
- Clone or copy this folder into your ComfyUI
custom_nodes/directory:custom_nodes/ └── nbpOpenNode-ComfyUI/ ├── __init__.py └── nbp_gemini_image2.py - Restart ComfyUI.
- The node will appear under api node → image → Gemini as NBP Gemini Image 2.
Requirements
No additional packages required — all dependencies ship with ComfyUI.
Authentication
This node uses the ComfyUI Cloud API. You must be logged in to your ComfyUI account within the application. The node will use your account's API key automatically; no separate Google or Vertex AI credentials are needed.