Burve List Google Image Models

The frustrating thing about Gemini image models is that they're API-only - you can't check a download folder to see what you own. Burve List Google Image Models is the read-only sanity check for that: it asks Google which image models your account actually has access to, and prints the result side by side with the plugin's own catalog so you can see what's available, what's retired, and where the mismatch is.

You want this node for two reasons. First, model availability genuinely varies by account, region, and provider - something that works for one GEMINI_API_KEY isn't guaranteed for yours, and Vertex access depends on your project and location. Second, Google retires things. Gemini 2.5 Flash Image, for example, is flagged for retirement on October 2, 2026, and preview models get renamed and normalized (the pack auto-migrates gemini-3.1-flash-image-preview to the stable ID). Before you build a workflow around a model, knowing its lifecycle status saves you from building on a deadline.

How it works

Two inputs, two providers. provider is aistudio or vertex. vertex_location_mode is auto (routes to global, where the stable 3.x image models live) or configured (uses your GOOGLE_CLOUD_LOCATION). For AI Studio it needs GEMINI_API_KEY in the environment; for Vertex it needs GOOGLE_CLOUD_PROJECT and GOOGLE_CLOUD_LOCATION. If the required variables are missing, it says so instead of crashing - that's the first thing you'll learn from this node, and honestly it's worth running for that alone.

On execution it calls client.models.list() on the Google Gen AI SDK, filters the results down to image models, and compares them against the pack's built-in catalog of the four it supports: gemini-3.1-flash-image, gemini-3-pro-image, gemini-3.1-flash-lite-image, and gemini-2.5-flash-image. The single model_report output is a JSON string with the provider/location, the discovered image models, and the catalog annotated with lifecycle status, retirement date, whether each was found in discovery, its API modes, and whether it's Vertex-global-only. API keys and credential contents are never included in the report.

Install

Pack-wide: ComfyUI Manager → search ComfyUI_Burve_Tools → install → restart, or

cd ComfyUI/custom_nodes
git clone https://github.com/Burve/ComfyUI_Burve_Tools.git
cd ComfyUI_Burve_Tools
pip install -r requirements.txt

Dependencies google-genai (≥2.13, <3), pillow, pillow-heif. ComfyUI 0.4.0+. Remember the environment-variable gotchas from the pack README: on Windows set the variable with setx and restart, on macOS standalone builds use launchctl setenv - GUI-launched ComfyUI doesn't reliably inherit your shell profile.

Gotchas

• It's a snapshot, not a promise. Discovery lists what the account can reach right now; quotas, billing, and moderation still apply per request.
• No environment variables → no report. The node returns a plain "not set" message rather than a guess, which is the correct behavior. Fix the env vars in the shell that launches ComfyUI, then restart.
• It needs an internet call to work. That's the point, but it also means it fails if you're offline.

It's a diagnostic node - you run it a few times while setting up, then forget it exists. For a small pack with no community lore to lean on, having a tool that tells you plainly whether your key works is more useful than it sounds.