[IMAGE] === COMFYUI BURVE TOOLS: GEMINI IMAGE NODES WITHOUT PUTTING YOUR API KEY IN THE WORKFLOW ===

Burve Tools is a collection of 19 ComfyUI nodes for Gemini image generation, reference handling, targeted edits, character planning, prompt tools, and workflow utilities.

I started building it after trying to use the Google image API at the end of last year. One custom node I found asked for the API key inside the node itself, which meant the key could be saved with the workflow metadata embedded in a generated image.

That felt like a very creative way to share an API key with the internet :D

I wanted the key to stay outside the workflow, so I asked Codex to help me build my own node. As often happens with vibe coding, one node somehow grew into an entire collection called ComfyUI Burve Tools.

The current version is 3.0.1.

--- THE TWO MAIN GENERATION NODES ---

Burve Google Image Gen uses Google AI Studio and reads GEMINI_API_KEY from the environment. The key is not an input on the node, so it is not stored inside the workflow.

Burve Google Image Gen (Vertex AI) uses a Google Cloud project, location, and Google credentials or Application Default Credentials. It does not use GEMINI_API_KEY.

Both nodes support Google's stable Gemini image models, including Gemini 3.1 Flash Image, Gemini 3 Pro Image, Gemini 3.1 Flash Lite Image, and the legacy Gemini 2.5 Flash Image. Available controls depend on the selected model and can include resolution up to 4K, aspect ratio, search, thinking, response format, and other model-specific options.

--- SOME USEFUL PARTS OF THE COLLECTION ---

- Reference packs: Bundle up to 14 images for Gemini. The newer Gemini Reference Pack can label each reference as generic, object, character, or style.

[NOTE] These nodes bundle reference images for the generator. They do not composite the images into one picture.

- Documents and video: Add PDF, TXT, Markdown, or CSV context, plus native video or a public YouTube URL when the selected model supports it.

- Crop and mask editing: Load an image, choose a fixed-ratio crop, then paint, erase, fill, inspect, or undo mask operations directly inside the node.

- Masked reinsertion: Generate a replacement for the cropped area and blend only the masked pixels back into the original image. Close aspect-ratio matches can be fitted automatically, while larger mismatches stop with an error instead of silently destroying the composition.

- Character planning: Build reusable adult character descriptions with body and appearance controls, face references, and optional fantasy-race details such as horns, wings, tails, scales, or claws.

- Prompt tools: Use system-instruction presets, a small prompt database, variable injection, and prompt selection helpers.

- Workflow utilities: Check AI Studio or Vertex authentication, inspect available Google image models, report image dimensions and aspect ratios, split images into grids, and save generated images in their original supported format.

The package also avoids writing API keys, prompts, uploaded media, or interaction contents to its plugin logs.

--- HOW TO INSTALL IT ---

Option 1: Open ComfyUI Manager and search for:

ComfyUI_Burve_Tools

Option 2: Install it from the public repository and follow the setup instructions in the README:

https://github.com/Burve/ComfyUI_Burve_Tools

The repository includes setup instructions for AI Studio and Vertex AI, plus debug nodes that show what authentication information ComfyUI can actually see. Remember to restart ComfyUI after changing environment variables.

This started as a small personal fix and became much larger than I originally planned. It has automated tests and more safeguards now, but I am not going to pretend that every possible ComfyUI setup and Gemini edge case has already been discovered.

Try version 3.0.1 and comment with the one fix, additional node, or video tutorial you would like me to work on next.

2 comments

[IMAGE] === COMFYUI BURVE TOOLS: GEMINI IMAGE NODES WITHOUT PUTTING YOUR API KEY IN THE WORKFLOW ===