Gemini Nano Banana 2.1 Arrives as an Official ComfyUI Partner Node

ComfyUI, the node-based visual workflow interface for generative AI, added support for Google's Gemini Nano Banana 2.1 as an official Partner Node. The integration brings multi-resolution image synthesis and text-driven image transformation directly into custom generation graphs, eliminating the need to bridge external scripts for Google's latest multimodal weights. By embedding these capabilities directly into the ComfyUI canvas, creators can chain reasoning-guided edits into broader production pipelines alongside existing diffusion models.

What's new

• Native Partner Node integration: Gemini Nano Banana 2.1 operates directly inside the ComfyUI workspace through dedicated partner nodes, allowing direct wiring into samplers, image loaders, and post-processing chains.
• Scalable resolution output: The model supports native generation from 1K up to 4K resolution, reducing the reliance on secondary upscalers for high-detail stills.
• Configurable thinking budgets: Users can toggle between Minimal, Medium, and High reasoning levels, balancing generation latency against prompt-adherence depth depending on task complexity.
• Instruction-based editing: The node accepts natural-language modification prompts to alter existing plates, change background elements, or swap subjects while maintaining composition.

How it fits your workflow

For storyboard artists, VFX concept designers, and art directors, this release makes direct image manipulation inside ComfyUI significantly faster. Instead of building multi-node ControlNet and inpainting masks for minor character or wardrobe tweaks, artists can route an image into the Gemini Nano Banana 2.1 node and specify precise instruction-based edits. The adjustable thinking parameters mean rapid low-compute ideation at Minimal thinking, followed by High-thinking passes when matching intricate set dressing or complex lighting setups.

In practical production, the model functions as a flexible alternative to traditional Stable Diffusion XL inpainting workflows and competes with instruction-tuned models like Flux.1 Fill and InstructPix2Pix. Because it handles 1K to 4K output natively, artists can generate high-resolution concept art without running secondary Latent Upscale or SUPIR nodes, saving GPU memory and simplifying graph layouts. ComfyUI pipelines can also chain Nano Banana 2.1 outputs directly into video generators like Kling 1.5, Runway Gen-3 Alpha, or AnimateDiff to create consistent first-frame anchors.

What it costs / how to try it

ComfyUI users can access Gemini Nano Banana 2.1 by updating their installation to the latest build and installing the official Partner Node package via the ComfyUI Manager. Running the node requires an active API configuration connected to Google's model endpoint, with usage billed according to standard developer platform rates based on token input, output resolution, and selected thinking tiers.