Google’s latest image model brings better editing, stronger references, and up to 4K output to ComfyUI.
AI image generation is moving quickly, and ComfyUI continues to make the latest models available to artists and creators. The newest addition is Gemini Nano Banana 2.1, now available in ComfyUI through Partner Nodes.
Nano Banana 2.1 builds on Nano Banana 2 with improvements in image quality, prompt understanding, text rendering, editing, and consistency. It also gives artists more control over how much processing the model uses before creating an image.
Better Image Quality
Nano Banana 2.1 supports 1K, 2K, and 4K output. Google says the updated model improves visual quality and realism across these resolutions.
It also fixes tiling problems that could appear in very wide or tall images. This is useful for artists creating banners, panoramic artwork, textures, and other unusual formats.
More Reference Images
One of the biggest improvements is reference-image support.
Nano Banana 2.1 can use up to 14 reference images in a single generation. Artists can combine characters, products, objects, backgrounds, and style references to create a more controlled result.
This can be especially useful for concept art, storyboards, product images, advertising, and character development.
Better Text Inside Images
Text generation has also received a major upgrade.
Nano Banana 2.1 is designed to create more accurate text, including signs, notices, numbers, and other written elements. This makes the model more useful for posters, advertisements, infographics, packaging concepts, and social media graphics.
For artists, this means fewer trips into Photoshop or another design application just to repair simple text errors.
Smarter Image Editing
Nano Banana 2.1 is not limited to creating images from scratch. It can make targeted changes to existing images using natural-language instructions.
You can ask it to change a color, replace a material, remove an object, or replace text while keeping the rest of the image intact. ComfyUI also provides workflows for global and mask-based editing.
This makes the model interesting for artists who want to iterate instead of starting over.
Three Thinking Levels
Another new feature is control over the model’s thinking level.
Creators can choose Minimal, Medium, or High thinking. Minimal is useful when exploring ideas quickly, while High can help with complex prompts and detailed compositions.
In practice, this gives artists a choice between speed and deeper processing.
Search-Grounded Images
Nano Banana 2.1 can also use Google Search and Image Search as grounding sources.
That can help when a prompt requires knowledge about a real landmark, unusual object, or other specific real-world subject.
A Strong New Tool for ComfyUI Artists
Nano Banana 2.1 is another sign that AI image tools are becoming more than simple text-to-image generators.
With stronger editing, reference images, text rendering, search grounding, flexible aspect ratios, and 4K output, the model gives digital artists more control over the creative process.
And because it is available through ComfyUI, artists can place it inside larger node-based workflows rather than treating image generation as a separate step.
For creators already using ComfyUI, Nano Banana 2.1 is worth exploring.
The banana just got a little smarter.