# Nano Banana Pro: A Practical Guide to ComfyUI API Workflows > Published 2026-04-02, updated 2026-09-05 · https://www.promptzone.com/seojun_zhao/nano-banana-pro-comfyui-node-streamlined-ai-art-creation-5738 --- title: "Nano Banana Pro: A Practical Guide to ComfyUI API Workflows" published: true description: "Nano Banana Pro in ComfyUI explained: load the official workflow, connect reference images, check credits, and understand hosted inference." tags: ai,comfyui,imagegeneration,gemini --- To use Nano Banana Pro in ComfyUI, open its official workflow template, sign in to a Comfy account with credits, and supply your prompt and any reference images. The Partner Node sends requests to a hosted service for Google's Gemini 3 Pro Image model. [ComfyUI tutorial][comfy], [Partner Nodes overview][overview], [Google announcement][launch] ## What are the key facts about Nano Banana Pro in ComfyUI? | Fact | Verified detail | | --- | --- | | Developer | Google DeepMind develops the model; ComfyUI documents its Partner Node integration. [Google announcement][launch], [ComfyUI tutorial][comfy] | | Released | Model announced November 20, 2025; a separate node release date is not published in the tutorial. [Announcement][launch], [tutorial][comfy] | | Type | Hosted image generation and editing reached through a ComfyUI Partner Node. [Tutorial][comfy] | | Size or parameters | Model parameter count not published; the node is a connector. [Model documentation][model], [tutorial][comfy] | | License and access | Closed model with no open weights; Partner Node calls require a Comfy account and credits. [Partner Nodes overview][overview] | | Where it runs | Model inference is remote; the workflow can be opened in local ComfyUI or Comfy Cloud. [Tutorial][comfy], [Partner Nodes overview][overview] | Your graph and the model have different execution locations. A local ComfyUI window can send inputs to a remote provider. Account access and the other nodes in your graph therefore deserve separate checks. [Partner Nodes overview][overview] ## What can the Nano Banana Pro ComfyUI workflow do? ComfyUI's official tutorial provides a ready-made Nano Banana Pro workflow and reference-image inputs. It documents image blending, text rendering, and subject consistency as model capabilities exposed through the integration. [ComfyUI tutorial][comfy] For practical work, use that template to keep a reference, instruction, and output in one visible workflow. Start with a simple composition task before adding other stages. This makes it easier to identify which stage needs attention when the result is wrong. Google's model documentation also lists thinking and search grounding as supported model capabilities. Check the node's exposed settings before assuming every Google API capability is available in the same form through ComfyUI. [Model documentation][model] Write a small acceptance brief for the graph. For a product composition, name the product features that must remain recognizable and the intended background. Use those criteria when comparing returned images, regardless of how many nodes the workflow contains. ## What does the Nano Banana Pro Partner Node require? Partner Nodes require a signed-in Comfy account, credits, and a supported network setup. The official overview explains login under Settings, User, and the credits controls under Settings, Credits. [Partner Nodes overview][overview] The provider receives the inputs used by its API node. A workflow containing a hosted node therefore does not keep all generation on your computer, even if you opened ComfyUI locally. [Partner Nodes overview][overview] Missing nodes can result from an outdated installation or a failed import. ComfyUI's tutorial warns that templates and node availability can differ between development and stable releases. [ComfyUI tutorial][comfy] Google's image guide says generated output counts may not exactly follow the request and that generated images contain SynthID. Inspect actual outputs instead of assuming the request's wording establishes their number or content. [Image guide][images] The model's documentation does not publish a parameter count or local weight package. Its hosted access provides no basis for a fixed local VRAM requirement or a guaranteed generation time for your entire graph. [Model documentation][model] ## How do you set up Nano Banana Pro in ComfyUI? ### Open the official template 1. Update ComfyUI using its documented installation path. If a required node is missing, check version compatibility and startup import messages. [ComfyUI tutorial][comfy] 2. Sign in through Settings, User, then inspect the balance under Settings, Credits. Confirm the account and network conditions in the Partner Nodes overview. [Overview][overview] 3. Open the Template Library and search for “Nano Banana Pro.” The official tutorial also provides a workflow download and a Comfy Cloud route. [Tutorial][comfy] 4. Supply your reference images to the template's `LoadImage` nodes, then enter the requested edit or composition. Inspect the available model settings before running. [Tutorial][comfy] Keep the template's first run focused. Use a reference of one object and request a different background while preserving the object's shape and colors. This is a suggested diagnostic task, not a claim about measured model performance. Examine the result before connecting more processing stages. If the request fails, first check whether the problem concerns account access, credits, connectivity, or node availability. Those are documented prerequisites for the hosted stage. [Overview][overview], [tutorial][comfy] If the request succeeds but the composition is wrong, revise the instruction and references. Treat that as an image-editing problem. Keeping request failures and visual failures separate makes troubleshooting more specific. ### Understand the direct API alternative You can also call the underlying model through Google's Gemini API. This uses a Google API key and Google's billing context, separately from ComfyUI Partner Node credits. [Google announcement][launch], [API key guide][keys], [Partner Nodes overview][overview] After creating a Gemini API key and setting `GEMINI_API_KEY`, this minimal request uses the current model identifier and documented API method. It is a direct Google request, not a command that executes the ComfyUI template. [Model documentation][model], [API reference][api] ```bash curl --fail-with-body \ "https://generativelanguage.googleapis.com/v1beta/models/gemini-3-pro-image:generateContent" \ -H "x-goog-api-key: $GEMINI_API_KEY" \ -H "Content-Type: application/json" \ -d '{"contents":[{"parts":[{"text":"Create a studio photograph of a plain red kettle on a cream background."}]}],"generationConfig":{"responseModalities":["TEXT","IMAGE"]}}' \ -o response.json ``` The file contains a JSON response envelope. Extract image content from returned parts and handle errors or missing images before passing the result into another stage. [API reference][api] Choose one access route for your initial evaluation and record it with the model name. This keeps account configuration and billing information understandable when someone else revisits the workflow. For ideas to adapt after the template works, see the sibling [Awesome Nano Banana Pro prompt collection guide](/ayaka_bui/awesome-nano-banana-pro-compact-ai-powerhouse-unveiled-5d7). ## How does Nano Banana Pro compare with Nano Banana 2? | Model | Documented distinction | | --- | --- | | Nano Banana Pro | Gemini 3 Pro Image focuses on complex visual design, mockups, and text-bearing output. [Model documentation][model] | | Nano Banana 2 | Gemini 3.1 Flash Image is presented as the speed-oriented counterpart for high-volume image work. [Nano Banana 2 documentation][flash] | Test your own references before choosing a model for a larger graph. Use the [ComfyUI complete guide](/tomas_novak/comfyui-2026-the-complete-guide-to-power-user-ai-image-generation-1g17) for workflow construction and node connections. ## How do you troubleshoot Nano Banana Pro in ComfyUI? ### Does the ComfyUI node download Nano Banana Pro? The Nano Banana Pro ComfyUI integration uses a Partner Node to access a closed hosted model; there are no open Nano Banana Pro weights to load locally. [Partner Nodes overview][overview] ### Do I need a Comfy account for the official Partner Node? The official Nano Banana Pro Partner Node requires a signed-in Comfy account and available credits. Review the supported network setup as well if login or execution fails. [Overview][overview] ### Can I use my Google API key in the example command? The Nano Banana Pro curl example calls Google's Gemini API directly with `GEMINI_API_KEY`. Its billing and authentication are separate from the ComfyUI Partner Node route described above. [API key guide][keys], [overview][overview] ### Why is the template missing a node? Missing Nano Banana Pro workflow nodes can result from ComfyUI version differences or failed startup imports, according to its tutorial. Check the official update guidance and import messages before replacing the node with an unrelated community extension. [Tutorial][comfy] ## Sources - [ComfyUI: Nano Banana Pro tutorial][comfy] - [ComfyUI: Partner Nodes overview][overview] - [Google: Introducing Nano Banana Pro][launch] - [Google: Gemini 3 Pro Image model documentation][model] - [Google: Image generation guide][images] - [Google: Using Gemini API keys][keys] - [Google: GenerateContent API reference][api] - [Google: Gemini 3.1 Flash Image model documentation][flash] [comfy]: https://docs.comfy.org/tutorials/partner-nodes/google/nano-banana-pro [overview]: https://docs.comfy.org/tutorials/partner-nodes/overview [launch]: https://blog.google/innovation-and-ai/products/nano-banana-pro/ [model]: https://ai.google.dev/gemini-api/docs/models/gemini-3-pro-image [images]: https://ai.google.dev/gemini-api/docs/image-generation [keys]: https://ai.google.dev/gemini-api/docs/api-key [api]: https://ai.google.dev/api/generate-content [flash]: https://ai.google.dev/gemini-api/docs/models/gemini-3.1-flash-image ## Related guides on PromptZone - [ComfyUI 2026: The Complete Guide](/tomas_novak/comfyui-2026-the-complete-guide-to-power-user-ai-image-generation-1g17) - [How to Install and Run SDXL Models in ComfyUI](/jaroslav/how-to-install-and-run-sdxl-models-in-comfyui-a-complete-guide-2nk2) - [How to Use LoRAs in ComfyUI in 2026](/tara_suzuki/how-to-use-loras-in-comfyui-in-2026-load-stack-and-troubleshoot-235e)