google / imagen4

Among the Google entries in the Civitai image pack, Imagen 4 is the quiet one. No search toggles like nano-banana-2, no 4K flagship billing like pro - just a clean text-to-image node with a negative prompt and aspect ratio, which makes it the most conventional Google node on the menu. If you've used a local SDXL workflow, this one will feel familiar in a way the Gemini-family nodes don't.

It's part of the official civitai/civitai-comfy-nodes pack: engine google, model imagen4, create-only (there's no edit variant). You type a prompt, it generates on Civitai's cloud, and an IMAGE comes back into your graph. No local model, no Google Cloud project, no Vertex AI billing account - your Civitai Buzz balance is the whole payment story.

How it works. Same pack machinery as always: the node builds a workflow payload, submits it to orchestration.civitai.com, and long-polls until the job lands. The interesting bit is what Imagen 4 exposes that the Gemini nodes don't: a real negative_prompt field. That's a genuinely useful control - being able to say "no text in the image, no watermark, no extra fingers" is something the Nano Banana nodes simply won't let you do, because Gemini doesn't take negatives. Imagen 4 also gives you aspect_ratio (1:1, 16:9, 9:16, 3:4, 4:3 - default 1:1), num_images (1-4), and a proper seed for reproducibility. output_format and image_metadata (external metadata stored with the image) round out the optional set.

Inputs that matter. prompt is the only required field. Beyond that: negative_prompt is your friend for cleaning up common artifacts, aspect_ratio picks the canvas (and unlike the width/height nodes, you're choosing from Imagen's native ratios rather than arbitrary pixels), and seed gives you repeatable runs. If you want to edit an existing image, this isn't the node - that's what the nano-banana nodes (or any other engine's edit variant) are for.

Install. Ships with the pack, so install once and the whole Civitai/Image/google menu appears:

cd ComfyUI/custom_nodes
git clone https://github.com/civitai/civitai-comfy-nodes.git
pip install -r civitai-comfy-nodes/requirements.txt   # requests

Or ComfyUI Manager → Custom Nodes Manager → search Civitai Comfy Nodes → Install, then restart. No model files to fetch.

Auth and gotchas. Credentials resolve from a wired Civitai Auth node, then CIVITAI_API_TOKEN, then a stored key or sidebar OAuth. Know your error codes: 402 = not enough Buzz, 401 = bad/expired token, 429 = rate limited. Jobs have a 30-minute timeout, and the pack is early preview - generated from a live API spec, so fields can shuffle between releases.

And the filter caveat is non-negotiable with anything Google: Imagen is a closed model with aggressive, opaque moderation, and Civitai applies its web-generator restrictions on top. Your prompt leaves the machine. If you want the closest thing to a local-style SDXL control set among Google's hosted options - negative prompt, aspect ratio, seed - this is the node to grab.