gemini-imagegen

name	gemini-imagegen
description	This skill should be used when generating and editing images using the Gemini API (Nano Banana Pro). It applies when creating images from text prompts, editing existing images, applying style transfers, generating logos with text, creating stickers, product mockups, or any image generation/manipulation task. Supports text-to-image, image editing, multi-turn refinement, and composition from multiple reference images.
model	sonnet

Gemini Image Generation (Nano Banana Pro)

Generate and edit images via Google's Gemini API. Requires GEMINI_API_KEY env var.

Model

Always use gemini-3-pro-image-preview (Pro, 1K-4K). Only switch models if explicitly requested.

Defaults

• Resolution: 1K — options 1K, 2K, 4K
• Aspect ratio: 1:1 — options 1:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 21:9

CRITICAL: File format

The API returns JPEG by default. Always save with .jpg:

image.save("output.jpg")   # ✅ correct
image.save("output.png")   # ❌ JPEG with PNG extension → media-type errors

Convert to PNG only if needed — see references/examples.md.

Code examples

All patterns (basic generation, custom resolution, editing, multi-turn, Search grounding, multi-reference composition, prompting recipes) in references/examples.md.

Minimal example:

import os
from google import genai
from google.genai import types

client = genai.Client(api_key=os.environ["GEMINI_API_KEY"])
response = client.models.generate_content(
    model="gemini-3-pro-image-preview",
    contents=["Your prompt"],
    config=types.GenerateContentConfig(response_modalities=['TEXT', 'IMAGE']),
)
for part in response.parts:
    if part.inline_data:
        part.as_image().save("output.jpg")

Prompting tips

• Photorealistic: include camera details (lens, lighting, angle, mood).
• Stylized art: specify style explicitly (kawaii, cel-shading, watercolor).
• Text in images: specify font style, placement.
• Product mockups: describe lighting setup and surface.
• Editing: describe changes conversationally — model handles semantic masking.

Notes

• All generated images include SynthID watermarks.
• Image-only mode (responseModalities: ["IMAGE"]) does not work with Google Search grounding.
• Default to 1K for speed; bump to 2K/4K when quality is critical.

Scripts

Helper CLI utilities (install dependencies with pip install -r requirements.txt):

• generate_image.py — text-to-image generation
• edit_image.py — edit an existing image with a text instruction
• compose_images.py — combine multiple reference images
• multi_turn_chat.py — interactive multi-turn refinement
• gemini_images.py — importable library wrapping the above