nano-banana

name	nano-banana
description	Generate or edit images using Nano Banana models (gemini-3.1-flash-lite-image, gemini-3.1-flash-image, gemini-3-pro-image, gemini-2.5-flash-image) with ADC auth. Supports "banana" as a verb (e.g., "please banana this") and Anime Dani / Chibi Dani character consistency.

Nano Banana Skill

Generate, edit, and iterate on visuals conversationally using Google's native Nano Banana image generation models (gemini-3.1-flash-lite-image, gemini-3.1-flash-image, gemini-3-pro-image, gemini-2.5-flash-image) with Application Default Credentials (ADC) authentication.

---

Trigger Conditions

Activate this skill whenever the user asks to:

• Generate new images or illustrations from text descriptions.
• Edit, transform, style, or combine existing images.
• Use "banana" as a verb (e.g., "please banana this image", "banana me as an anime character", "banana this into chibi style").
• Generate character-consistent imagery featuring Anime Dani, Daniela, or Chibi Dani.

---

Model Selection & Capability Matrix

Feature / Capability	Nano Banana 2 Lite (nano-banana-2-lite)	Nano Banana 2 (nano-banana-2)	Nano Banana Pro (nano-banana-pro)	Nano Banana Legacy (nano-banana)
Model ID	gemini-3.1-flash-lite-image	gemini-3.1-flash-image	gemini-3-pro-image	gemini-2.5-flash-image
Primary Use Case	High-velocity, budget-sensitive tasks	Generalist workhorse balance of speed & quality	Premium professional asset production	Legacy workhorse (transition to 2 Lite)
Supported Resolutions	1K only	512px (0.5K), 1K, 2K, 4K	1K, 2K, 4K	1024px (1K)
Object References	Up to 14	Up to 10	Up to 6	Up to 3
Character References	N/A	Up to 4	Up to 5	N/A
Style References	N/A	N/A	Up to 3	N/A
Total Ref Images	Up to 14	Up to 14	Up to 14	Up to 3
Search Grounding	Not Supported	Web Search + Image Search	Web Search	Not Supported
Thinking Mode	Standard	minimal / high levels	Default thinking	N/A
Video Context	Not Supported	YouTube URLs & MP4 files	Not Supported	Not Supported
Interleaved Output	No	No	Supported (text + images)	No

Model Selection Guide

• Nano Banana 2 Lite (nano-banana-2-lite or nano-banana-lite): Default model. Ultra-fast, budget-conscious, high-volume generation (gemini-3.1-flash-lite-image). Ideal for rapid prototyping and high-velocity iteration.
• Nano Banana 2 (nano-banana-2): The versatile, all-around workhorse (gemini-3.1-flash-image). Best for general tasks, multi-reference object/character composition, video-to-image, 4K output, and image search grounding.
• Nano Banana Pro (nano-banana-pro): Premium model (gemini-3.1-pro-image). Use for complex visual reasoning, intricate text rendering, professional marketing assets, character consistency with style references, and interleaved story illustrations.
• Nano Banana (nano-banana): Legacy pioneer model (gemini-2.5-flash-image). Transition users to Nano Banana 2 Lite for superior quality, lower cost, and lower latency.

---

Technical Specifications & Configuration

Resolution & Formatting Rules

• Resolution Parameter (image_size): Must use uppercase 'K' (e.g. 512px, 1K, 2K, 4K). Lowercase values (e.g. 1k) will be rejected.
• SynthID Watermark: All generated images automatically include an imperceptible SynthID watermark.

Supported Aspect Ratios

• Nano Banana 2 Lite: 1:1, 3:2, 2:3, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 21:9.
• Nano Banana 2: 1:1, 1:4, 1:8, 2:3, 3:2, 3:4, 4:1, 4:3, 4:5, 5:4, 8:1, 9:16, 16:9, 21:9.
• Nano Banana Pro: 1:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 21:9.
• Nano Banana Legacy: 1:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 21:9.

---

Character Consistency (Anime Dani, Daniela & Chibi Dani)

When generating or editing imagery featuring Anime Dani, Daniela, or Chibi Dani, attach the corresponding reference images from {baseDir}/references/ to enforce visual consistency.

Anime Dani / Daniela

• Style: Vibrant anime aesthetics, expressive eyes, crisp cel-shading, detailed linework.
• Reference Images: Pass {baseDir}/references/speaker-dani.png and {baseDir}/references/celebrating-dani.png via -i.

Chibi Dani

• Style: Super-deformed (SD) kawaii chibi style, oversized head, cute proportions, simplified outlines.
• Reference Image: Pass {baseDir}/references/chibi-dani.png via -i.

---

Prompting Strategies & Templates

Text-to-Image Generation

1. Photorealistic Scenes

• Template: A photorealistic [type of shot] of a [subject description] in a [setting description]. [Lighting description]. Shot from a [camera angle] with a [lens type]. [Aspect ratio].
• Example: A photorealistic wide-angle shot of a vibrant coral reef teeming with tropical fish. Crystal-clear turquoise water with sunbeams filtering down from the surface, illuminating a sea turtle gliding gracefully over the coral. Shot from a low perspective with a wide-angle lens. Aspect ratio 16:9.

2. Stylized Illustrations & Stickers

• Template: A [style] of a [subject with details] doing [activity]. Features [bold outlines, cel-shading, etc.] and [color/background preference].
• Example: A kawaii-style sticker of a happy red panda wearing a tiny bamboo hat, munching on a green bamboo leaf. Bold clean outlines, simple cel-shading, vibrant colors. Solid white background.

3. Accurate Text Rendering

• Template: Create a [image type] for [brand/concept] with the text "[exact text]" in a [font style]. The design should be [style description], with a [color scheme].
• Example: Create a modern minimalist logo for a coffee shop called 'The Daily Grind'. The text should be in a clean bold sans-serif font. Black and white color scheme inside a circle, incorporating a coffee bean.

4. Product Mockups & Commercial Photography

• Template: A high-resolution studio-lit product photograph of a [product] on a [surface/background]. Lighting is [lighting setup] to [purpose]. Camera angle is [angle] with sharp focus on [key detail].
• Example: A high-resolution studio-lit product photo of a minimalist matte black ceramic coffee mug on polished concrete. Three-point softbox lighting, soft diffused highlights. 45-degree angle shot with sharp focus on rising steam.

5. Minimalist & Negative Space Design

• Template: A minimalist composition featuring a single [subject] positioned in the [corner/side]. The background is a vast, empty [color] canvas with negative space. Soft lighting.
• Example: A minimalist composition featuring a delicate red maple leaf in the bottom-right frame. Vast empty off-white canvas with negative space for text overlay. Soft diffused light.

6. Sequential Art & Storyboarding

• Template: Make a [N]-panel comic in a [art style] featuring [character/setting].
• Example: Make a 3-panel comic in a gritty noir art style with high-contrast black and white inks showing a humorous detective scene.

7. Grounded Search Generation

• Usage: Enable google_search tool to pull real-time weather, sporting events, or live data into visual graphics.
• Example: Visualize the current 5-day weather forecast for San Francisco as a clean modern weather chart with attire suggestions for each day.

---

Image Editing & Multimodal Transformations

1. Adding and Removing Elements

• Template: Using the provided image of [subject], please [add/remove/modify] [element]. Ensure the change [integration description].
• Example: Using the provided image of my cat, please add a small knitted wizard hat on its head. Ensure it sits comfortably and matches the soft lighting.

2. Inpainting (Semantic Masking)

• Template: Using the provided image, change only [specific element] to [new element]. Keep everything else in the image exactly the same, preserving original style and lighting.
• Example: Using the provided image of a living room, change only the blue sofa to a vintage brown leather chesterfield sofa. Keep all other room elements unchanged.

3. Style Transfer

• Template: Transform the provided photograph of [subject] into the artistic style of [artist/style]. Preserve the original composition but render with [stylistic details].
• Example: Transform the provided photograph of a modern city street into the style of Van Gogh's Starry Night with swirling impasto brushstrokes and deep blue/yellow palette.

4. Advanced Multi-Image Composition

• Template: Create a composite image combining elements from the provided images. Take [element from Image 1] and place it on/with [element from Image 2].
• Example: Take the blue floral dress from the first image and have the woman from the second image wear it in an outdoor e-commerce fashion shot.

5. Sketch-to-Photo (Bringing Drawings to Life)

• Template: Turn this rough [medium] sketch of a [subject] into a [style] photo. Keep [sketch lines/profile] but add [materials/lighting].
• Example: Turn this rough pencil sketch of a futuristic car into a polished photo of a concept car in a showroom with metallic blue paint and neon rim lighting.

---

Professional Prompting Best Practices

1. Be Hyper-Specific: Describe materials, lighting setups, camera lenses, and exact textures (e.g. "ornate elven plate armor etched with silver leaf patterns, high pauldrons shaped like falcon wings").
2. Specify Intent & Context: Framing the purpose improves composition (e.g. "e-commerce product shot for high-end skincare" vs "bottle photo").
3. Iterate Conversationally: Use multi-turn follow-ups to fine-tune details ("Keep everything the same, but make the background light warmer").
4. Positive Framing for Negative Prompts: Instead of saying "no cars", describe the scene positively as "an empty, deserted street with no signs of traffic".
5. Cinematic Camera Controls: Use photographic terms like wide-angle shot, macro shot, low-angle perspective, three-point softbox setup.

---

Code Integration Guide (Google GenAI SDK)

Python SDK (google-genai)

from google import genai
import base64

client = genai.Client()

# Text-to-Image Generation
interaction = client.interactions.create(
    model="gemini-3.1-flash-image",
    input="A futuristic city inside a floating glass bottle in space",
    generation_config={"thinking_level": "high"},
    response_format={
        "type": "image",
        "aspect_ratio": "16:9",
        "image_size": "2K"
    }
)

with open("output.png", "wb") as f:
    f.write(base64.b64decode(interaction.output_image.data))

Multimodal Image Editing with Base64

with open("input.png", "rb") as f:
    img_bytes = f.read()

interaction = client.interactions.create(
    model="gemini-3.1-flash-image",
    input=[
        {"type": "text", "text": "Add a glowing blue wizard hat to the cat"},
        {
            "type": "image",
            "data": base64.b64encode(img_bytes).decode('utf-8'),
            "mime_type": "image/png"
        }
    ]
)

---

Usage & Execution via CLI

Note: {baseDir} refers to the root directory of the installed skill.

Use uv run to execute the image generation script at {baseDir}/scripts/banana.py.

CLI Arguments

• -p, --prompt: Text prompt describing the image to generate or edit (required).
• -f, --filename: Output image filename (required).
• -i, --input-image: Path to input/reference image(s). Specified up to 14 times.
• -m, --model: Model selection (nano-banana-2-lite, nano-banana-2, nano-banana-pro, nano-banana). Default: nano-banana-pro.
• -r, --resolution: Output resolution (1K, 2K, 4K). Default: 1K.
• --project: GCP Project ID for Vertex AI (optional).
• --location: GCP Location for Vertex AI (optional, default: us-central1).

Examples

# Ultra-fast generation with Nano Banana 2 Lite
uv run {baseDir}/scripts/banana.py -p "A minimalist tech workspace with a laptop and coffee cup" -f "workspace_lite.png" -m "nano-banana-2-lite"

# High-resolution 4K generation with Nano Banana 2
uv run {baseDir}/scripts/banana.py -p "A photorealistic macro shot of a Monarch butterfly on a flower" -f "butterfly_4k.png" -m "nano-banana-2" -r "4K"

# Character Consistency Generation with Nano Banana Pro
uv run {baseDir}/scripts/banana.py -p "Anime Dani presenting a tech keynote speech" -f "dani_keynote.png" -i "{baseDir}/references/speaker-dani.png" -i "{baseDir}/references/celebrating-dani.png" -m "nano-banana-pro"

# Image Editing ("banana this")
uv run {baseDir}/scripts/banana.py -p "banana this photo into chibi style" -i "photo.png" -i "{baseDir}/references/chibi-dani.png" -f "chibi_output.png" -m "nano-banana-2"