app.py

import gradio as gr
from google import genai
from google.genai import types
import os
from PIL import Image
import io

API_KEY = os.environ.get("GOOGLE_API_KEY")
client = genai.Client(api_key=API_KEY)

def generate_image(description: str, style: str) -> Image.Image | None:
    """Generates an image using Gemini 2.0 Flash based on description and style."""

    if not description:
        print("Description is empty. Cannot generate image.")
        return None

    # Construct the prompt for the image generation model
    image_prompt = f"""
    Please create an illustration based on the following description and style.
    - Description: {description}
    - Style: {style}
    - Format: High resolution, detailed image.
    - Aspect Ratio: Square (1:1) is preferred, but follow the description if it implies a different ratio.
    - **DO NOT** include any text in the image.
    """

    print(f"Generating image with prompt: {image_prompt}")

    try:
        # Call the Gemini API
        image_response = client.models.generate_content(
            model="gemini-2.0-flash-preview-image-generation",
            contents=image_prompt,
            config=types.GenerateContentConfig(
                response_modalities=['TEXT', 'IMAGE']
            )
        )

        # Process the response to extract the image
        if image_response and image_response.candidates and image_response.candidates[0].content.parts:
            for part in image_response.candidates[0].content.parts:
                if part.inline_data and part.inline_data.mime_type.startswith('image/'):
                    image_data_base64 = part.inline_data.data
                    img = Image.open(io.BytesIO(image_data_base64))

                    img = img.resize((512, 512), Image.Resampling.LANCZOS)
                    print("Image generated successfully.")
                    return img

            print("Image data not found in any response parts or not an image.")
            return None
        else:
            print("Image generation response is empty or malformed.")
            return None

    except Exception as e:
        print(f"Error generating image: {e}")
        # Handle specific API errors if needed
        return None

# Define available styles for the dropdown
styles = [
    "Sketch - 草圖",
    "Watercolor - 水彩",
    "Oil Painting - 油畫",
    "Digital Art - 數位藝術",
    "Cartoon - 卡通",
    "Photorealistic - 寫實",
    "Abstract - 抽象派",
    "Pixel Art - 像素藝術",
    "Anime - 動畫",
    "Impressionist - 印象派"
]

# Create the Gradio interface
app = gr.Interface(
    fn=generate_image,
    inputs=[
        gr.Textbox(label="✏️ Enter image description", lines=3, interactive=True),
        gr.Dropdown(choices=styles, label="🎨 Select Style", value="Sketch - 草圖") # Default style
    ],
    outputs=gr.Image(label="Generated Image"),
    title="🤖 Talk to Draw！ 🎨",
    description="Collaborate with AI to draw images from text or voice descriptions with different styles, powered by Google Gemini 2.0 Flash.",
    flagging_mode="never" # Disable flagging feature
)

app.launch()