Image Params (OpenRouter)

"Image Params (OpenRouter)" shapes the images that Image (OpenRouter) generates, and it's the config node with the most toys. It takes the base_config from Base Config (OpenRouter) and adds aspect ratio, image size, and temperature to produce the OR_IMAGE_CONFIG the Image (OpenRouter) node runs on. Purely data - no API call, no generation. Just the middle of Base Config (OpenRouter) → Image Params (OpenRouter) → Image (OpenRouter) → picture.

The inputs that matter

• aspect_ratio - ten presets, and this is where the OpenRouter variant flexes: 1:1, 16:9, 4:3, 9:16, 3:4, 2:3, 3:2, 4:5, 5:4, and the ultrawide outlier 21:9. The LiteLLM side of the pack only offers five. That's not an accident - OpenRouter passes Gemini's fuller set of ratios through, so this is the node to reach for when you need something less standard than square-or-landscape.
• image_size - 1K, 2K, or 4K, i.e. roughly 1M, 4M, or 16M pixels. The same cost-and-speed dial as elsewhere in the pack: 1K to iterate, 2K for real output, 4K when you want the megapixels and are willing to wait - the Image (OpenRouter) node gives 4K requests a 10-minute timeout because they genuinely can take that long.
• temperature - 0 to 1, default 1.0. The README's guidance holds here: 0 for deterministic, 0.5 as the balanced default, 1.0 for creative range. Image generation caps at 1, unlike chat's 2.

The single output is config of type OR_IMAGE_CONFIG, wiring into the Image (OpenRouter) node. One behind-the-scenes note: on the OpenRouter side, the Image node builds its image_config from whatever aspect_ratio and image_size it finds in the config - so this node's values genuinely control what gets sent to the API.

Installing it

# ComfyUI Manager: search "Gemini LiteLLM", Install, restart

# or:
cd ComfyUI/custom_nodes
git clone https://github.com/ZUENS2020/ComfyUI-Gemini-LiteLLM

Zero-dependency pack: no requirements.txt, no pip install, no model files.

Gotchas

Keep it in the right half of the pack - OR_IMAGE_CONFIG only connects to the OpenRouter pipeline, and mixing it with the LiteLLM nodes' types will get you refused connections. And remember the cost reality of the dials: every extra ratio or 4K request is real spend on your OpenRouter account, so test cheap before you commit to big.