⭐ Star Nano Banana (Gemini Image Gen)
The node that calls Google (bring a key)
- image1
- image2
- image3
- image4
- image5
- image
- prompt
Let's get the two things that will save you time right away. First: "Nano Banana" is not a model you download - it's Google's Gemini image model, and this node is a ComfyUI front-end for the API. Second: the name is a lie in the friendliest possible way, because the node does call an API and you will need a key. If you searched for this hoping for a local model, stop here and read the Qwen-Image-Edit article instead - that one is Apache 2.0 and free.
StarNanoBanana is a text-to-image and image-editing node pointed at gemini-2.5-flash-image-preview, the original consumer Nano Banana. You give it a prompt, optionally one to five reference images, and it hands you back a finished IMAGE plus the prompt string it actually sent. No GPU work on your side, no checkpoint files, no VAE dance. The trade is the usual one for closed models: it's watermarked with invisible SynthID, billed per image, and Google decides a lot of the details.
What you actually set
- prompt - your instruction, with a decent "cinematic fantasy landscape" default so the node works the moment you add it.
- ratio - 1:1, 16:9, 9:16, 4:3, 3:4. Standard stuff.
- megapixels - 1 MP up to 12 MP. This is the resolution dial. At 1 MP you're getting roughly 1024×1024; push it when you want a bigger canvas and are fine paying for it.
- prompt_template - the sleeper hit. Thirty built-in edit prompts (style transfer, color enhancement, background change, object removal, face enhancement, and more). Pick one and it overrides the prompt box with a carefully-worded instruction. This is the closest the node comes to a "just make it good" button.
- image1–image5 - optional reference images for editing or style-matching. This is where the node shines: it's not just T2I, it's "edit this photo in the way Nano Banana edits photos."
Wiring it up
Outputs are image (wire it into a Save Image or a preview) and prompt (handy if you're logging what actually went out, or feeding it into metadata). Nothing else needed - it's a self-contained terminal node.
Install
This pack's README is worth reading once because it's the whole story: ComfyUI_StarBetaNodes is the beta repo, and there are no active nodes left in it - everything moved to the main ComfyUI_StarNodes repo. So install that one, not this one:
- ComfyUI Manager → search
Starnodes→ install → restart, or cd ComfyUI/custom_nodes && git clone https://github.com/Starnodes2024/ComfyUI_StarNodesthen restart.
If you already installed the beta repo, remove it. The author, Starnodes2024, is a solo dev who also ships a Windows Image Manager app - friendly, active, and this pack is pure Python, so there's no heavy dependency hunt.
Where people get burned
The API key. You'll need one from Google AI Studio (the Gemini API), and the free tier runs out faster than you think at 4+ MP. Expect 401/403 errors if the key is missing or mistyped, and 429 rate-limit noise when the free tier is exhausted. And remember the watermark: every image carries an invisible SynthID. If you need license-clean, local, no-key editing, the honest advice is to reach for Qwen-Image-Edit instead - the node makes the banana convenient, not free.
Inputs (10)
| Name | Type | Default | Description |
|---|---|---|---|
| prompt | STRING | A beautiful fantasy landscape, cinematic lighting | — |
| model | COMBO | gemini-2.5-flash-image-preview | 1 options: gemini-2.5-flash-image-preview |
| ratio | COMBO | 1:1 | 5 options: 1:1, 16:9, 9:16, 4:3, 3:4 |
| megapixels | COMBO | 1 MP (≈1024x1024) | 15 options: 1 MP (≈1024x1024), 2 MP (≈1448x1448), 3 MP (≈1774x1774), 4 MP (≈2048x2048), 5 MP (≈2289x2289), 6 MP (≈2508x2508), +9 |
| prompt_templateopt | COMBO | Use Own Prompt | 30 options: Use Own Prompt, Style Transfer: Transform the image to impressionist painting style with visible brush strokes and vibrant colors, Color Enhancement: Enhance colors, increase saturation, improve contrast, make image more vibrant and appealing, Background Change: Replace the background with a studio setup, clean white backdrop, professional lighting, Object Removal: Remove the distracting objects in background, keep only the main subject, clean composition, Lighting Adjustment: Brighten the image, improve lighting, add dramatic shadows, enhance overall illumination, +24 |
| image1opt | IMAGE | — | |
| image2opt | IMAGE | — | |
| image3opt | IMAGE | — | |
| image4opt | IMAGE | — | |
| image5opt | IMAGE | — |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| image | IMAGE | — |
| prompt | STRING | — |