Nano Banana uses the Google Gemini API which requires a Google AI Studio API key and billing enabled. The cost is approximately $0.04 per image generated, making it affordable for most use cases.

How do I keep a character consistent with Nano Banana?

Upload a reference image of your character into the Nano Banana ComfyUI node. The model uses that reference as its anchor for every generation — keeping face, hair, and key features consistent while applying your new prompt changes.

How do I turn Nano Banana images into videos?

Use the WAN 2.2 image-to-video workflow in ComfyUI. Upload your consistent character images from Nano Banana, set your prompt, output dimensions and frame count, then run the workflow to generate a cinematic video.

How to Create Consistent Characters with Nano Banana in ComfyUI (2026)

Introduction

What you'll learn: How to generate a consistent AI character using Nano Banana in ComfyUI, edit it with targeted prompts, blend multiple reference images, and animate the result into a cinematic video using WAN 2.2.

If you've ever wanted to create an AI character that looks the same in every image, Nano Banana makes it easy. In this guide, you'll learn how to use the Nano Banana ComfyUI workflow to design consistent characters, edit them with precision, and even turn them into cinematic videos using WAN 2.2. By the end, you'll know how to create, refine, and animate your own AI characters from start to finish.

Here's an example — a cinematic video created using just one reference image with Nano Banana, showing consistent character appearance across every frame, animated with WAN 2.2:

Reference image used

Single reference image of a red-haired woman used as input for the Nano Banana consistent character workflow

Generated cinematic video

Cinematic desert bike video generated from one reference image using Nano Banana + WAN 2.2

What is Nano Banana?

Nano Banana is an AI image generation and editing model that runs via the Google Gemini API. It's designed specifically for consistent character generation — letting you change styles, outfits, backgrounds, poses, and expressions while keeping your character's core appearance the same across every image.

This makes it particularly useful for anyone building characters for comics, animation, digital storytelling, social media content, or AI video workflows. Unlike standard image generation where each output looks different, Nano Banana treats your reference image as a persistent identity anchor.

🎨

Consistent identity

Same face, hair, and features across every generation — regardless of pose or setting.

✏️

Prompt-based edits

Add, remove, or change elements with simple text instructions.

🖼️

Multi-reference blending

Combine 2–3 reference images to create complex, consistent composite characters.

Setting Up the Nano Banana ComfyUI Workflow

Step 1 — Download the Workflow

Start by downloading the Nano Banana ComfyUI workflow and opening it in ComfyUI. Make sure ComfyUI is updated to the latest version. If there are any missing nodes after loading, use the ComfyUI Manager to install them, then restart ComfyUI.

Download Nano Banana Workflow

💡 Tip: New to ComfyUI? How to Install ComfyUI will get you set up from scratch.

Nano Banana workflow loaded in ComfyUI showing all nodes connected correctly — Nano Banana workflow loaded and ready in ComfyUI

Step 2 — Get Your Nano Banana API Key

Nano Banana runs on Google Cloud via the Gemini API, so you'll need a Google AI Studio API key with billing enabled. The cost is approximately $0.04 per image — very affordable for the quality you get.

Visit aistudio.google.com
Go to Dashboard → API Keys → Create API Key.
Copy your new API key.
Set up billing in your Google account (required for Nano Banana).
Paste your API key into the Nano Banana node in ComfyUI when prompted.

Google AI Studio interface showing the API Keys section and Create API Key button — Create your API key in Google AI Studio — Dashboard → API Keys → Create API Key

Generating Consistent Character Images

With the workflow open and your API key in place, you're ready to start generating. Here's how the generation process works:

1. Upload a Reference Image

Upload a reference image of your character into the image input node. This image becomes the identity anchor — Nano Banana will keep the core appearance of this character in every generation.

💡 Tip: Don't have a reference image yet? Use the Qwen text-to-image workflow to generate your first character reference image.

Reference image of a red-haired girl uploaded into the Nano Banana ComfyUI image input node — Upload your reference image into the image input node — this becomes your character's identity anchor

2. Configure the Model and Prompts

In the Nano Banana node, select the gemini-2.5-flash-image-preview model. You'll find two prompt fields:

Prompt 1: Describe the character and scene — for example: "girl in red dress standing in a forest, cinematic lighting".
Prompt 2: Specify camera angle or visual style — for example: "cinematic close-up, soft lighting, shallow depth of field".

💡 Tip: Use ChatGPT or Claude to generate styling prompts. Share your character description and ask for variations: "Give me 10 different scene and outfit prompts for a red-haired action hero."

Nano Banana ComfyUI node showing the prompt fields and gemini-2.5-flash-image-preview model selected — Set the model to gemini-2.5-flash-image-preview and fill in both prompt fields

3. Run and Review Output

Once your prompts are ready, click Queue Prompt to run the workflow. You'll get a new, consistent image with your character's appearance preserved and the scene changed according to your prompt.

Nano Banana output — red-haired woman running through a desert with a sandstorm forming behind her, character features consistent with reference — Example output — same character, new scene. The face, hair colour, and core features remain consistent.

Editing and Refining Your Character

Nano Banana's editing mode lets you make targeted changes to any generated image using simple text instructions. Feed any output image back in as the new reference and describe what you want to change.

Edit 1 — Remove an Object

Use a prompt to remove an object or element from the image while keeping everything else consistent:

Original Nano Banana output showing the red-haired girl with an object in her hand — before editing

Edited Nano Banana output with the object removed from the girl's hand — she runs naturally with arms free

Prompt: "Remove the object from her hands and make her look like she's running naturally — a determined young woman running across a hot desert under the afternoon sun, arms moving freely, sand swirling around her boots, hair blowing in the wind, cinematic action shot."

Edit 2 — Change Gaze Direction

Adjust where the character is looking with a direct instruction:

Nano Banana edit output — red-haired girl now looking behind her instead of toward the camera, gaze direction changed — After edit: character now looks behind her — gaze direction changed with a single prompt instruction

Prompt: "Make the girl look behind her and not toward the camera."

Edit 3 — Add a Background Element

Add new environmental elements to the scene without altering the character:

Nano Banana edit output — massive sandstorm added to the background behind the running red-haired girl, dramatic cinematic atmosphere — After edit: a massive sandstorm added to the background — same character, dramatically new atmosphere

Prompt: "Make a massive sandstorm forming behind her — a red-haired woman running through the desert, looking back in fear, wind whipping her hair and clothes, sunlight dimmed by swirling sand, cinematic dramatic scene."

This iterative editing process is what makes Nano Banana powerful for character-based workflows — you can refine your character's look across many scenes, each building naturally on the last.

Using Multiple Reference Images

Nano Banana also lets you combine two or three reference images to generate a single new consistent image. Connect the Multi Image node to the Nano Banana node, enable both Image Input and Multi Image, then upload your reference images. The model will blend key features from all provided references.

For example: to show a girl riding a bike, use a reference image of the girl and a separate reference of the bike. With the right prompt, Nano Banana blends both into one consistent output.

Reference images

Reference image 1 — red-haired girl character for multi-reference Nano Banana input — Reference 1: character

Reference image 2 — bicycle used as the second reference for multi-reference Nano Banana blending — Reference 2: bike

Output image

Nano Banana multi-reference output — red-haired girl riding a bike, blending both reference images into one consistent result — Multi-reference output: girl riding the bike — features blended from both references into one consistent image

Creating Cinematic Videos with WAN 2.2

Once you've designed your consistent character images with Nano Banana, use the WAN 2.2 image-to-video workflow to animate them into smooth, cinematic videos. WAN 2.2 takes your static images and generates motion that matches your prompt.

Download WAN 2.2 Image-to-Video Workflow

Download and open the WAN 2.2 image-to-video workflow in ComfyUI.
Install any missing models or nodes shown in the ComfyUI Manager.
Upload your Nano Banana character images into the image input nodes.
Set your output width, height, and frame count.
Enter your motion prompt, then run the workflow.

Images used for video

Input images used for the WAN 2.2 image-to-video workflow — consistent character frames from Nano Banana

Generated video

WAN 2.2 output — cinematic desert scene with sandstorm, generated from Nano Banana character images

First and Last Frame Connection (WAN 2.2 FALI Workflow)

To create smooth, seamless transitions between video clips, use the FALI workflow (First And Last Image). This workflow takes the first frame of one video and the last frame of another, then generates smooth in-between motion — making your final cut look continuous and cinematic.

Download WAN 2.2 FALI (Connection) Workflow

First and last frame images

First frame image — red-haired girl hopping onto a bike in the desert, used as the start frame for the FALI workflow — First frame

Last frame image — red-haired girl riding the bike through a desert with a sandstorm behind her, used as the end frame for FALI — Last frame

Generated connection video

FALI workflow output — smooth cinematic transition generated between the first and last frame images

Conclusion

The Nano Banana ComfyUI workflow is a simple, powerful way to create and iterate on AI-generated consistent characters. Whether you're building a character for a comic, a social media series, or a full cinematic video, the combination of Nano Banana for image consistency and WAN 2.2 for animation gives you a complete creation pipeline inside ComfyUI.

One reference image is enough to anchor consistent character identity across unlimited generations.
Editing is targeted and iterative — each prompt change builds on the last without losing character features.
Multi-reference blending lets you add props, vehicles, or environments while keeping character consistency.
WAN 2.2 + FALI turns your static character frames into smooth, seamless cinematic clips.

Ready to start? Download all three workflows above, grab your Google API key, and follow this guide step by step. If you run into node errors, the ComfyUI Troubleshooting Guide covers the most common fixes.

Frequently Asked Questions

Nano Banana is an AI image generation and editing model that runs via the Google Gemini API (Google Cloud). It specialises in consistent character generation — keeping a character's face, hair, and core features the same across every image while applying your prompt-based scene and style changes.

Nano Banana uses the Google Gemini API which requires a Google AI Studio API key with billing enabled. The cost is approximately $0.04 per image generated — very affordable for the quality and consistency it produces.

Upload a reference image of your character into the Nano Banana ComfyUI image input node. The model uses that reference as its identity anchor for every generation — keeping the face, hair colour, and key features consistent while applying your new scene or style prompt.

Yes. Connect the Multi Image node to the Nano Banana node, enable both Image Input and Multi Image, then add two or three reference images. The model will blend key features from all references into a single consistent output — useful for adding props, vehicles, or different character elements.

Use the WAN 2.2 image-to-video workflow in ComfyUI. Upload your consistent character images from Nano Banana, set your motion prompt, output dimensions, and frame count, then run the workflow. The FALI workflow can then be used to create smooth transitions between separate video clips.