ComfyUI-Gemini

Why we included this project

Google's Gemini models plug into ComfyUI through this custom node pack, which adds nodes for generating prompts, describing images, and running multi-turn chat. It covers the text-only Gemini Pro, the vision model, and the newer 1.5 Pro, which adds system instructions and single-file uploads for audio, PDFs, and text. The batch-labeling workflow is useful if you train LoRAs, since Gemini Pro Vision can tag images automatically. Because the API key can be set as an environment variable, shared workflows don't leak credentials. It's a clean way to bring a capable external model into an existing image-generation setup.