index.py

import os
import base64
import numpy as np
from PIL import Image, ImageChops, ImageDraw

import io
import requests
import replicate
import gradio as gr

from dotenv import load_dotenv, find_dotenv

# Locate the .env file
dotenv_path = find_dotenv()
load_dotenv(dotenv_path)
REPLICATE_API_TOKEN = os.getenv('REPLICATE_API_TOKEN')

def generate_sd3_image(prompt):
    prompt = "texture sample zoomed in of " + prompt  + ", perfect, seamless pattern"

    output = replicate.run(
        "stability-ai/stable-diffusion-3",
        input={
            "cfg": 3.5,
            "steps": 28,
            "prompt": prompt,
            "aspect_ratio": "1:1",
            "output_format": "jpg",
            "output_quality": 80,
            "negative_prompt": "",
            "prompt_strength": 0.85
        }
    )

    response = requests.get(output[0])

    return Image.open(io.BytesIO(response.content))

def generate_pattern(image, prompt):

    if image is not None:
        # Convert the numpy array to a PIL image
        starter_image_pil = Image.fromarray(image.astype('uint8'))

        # Resize the starter image if either dimension is larger than 768 pixels
        if starter_image_pil.size[0] > 768 or starter_image_pil.size[1] > 768:
            # Calculate the new size while maintaining the aspect ratio
            if starter_image_pil.size[0] > starter_image_pil.size[1]:
                # Width is larger than height
                new_width = 768
                new_height = int((768 / starter_image_pil.size[0]) * starter_image_pil.size[1])
            else:
                # Height is larger than width
                new_height = 768
                new_width = int((768 / starter_image_pil.size[1]) * starter_image_pil.size[0])
            
            # Resize the image
            starter_image_pil = starter_image_pil.resize((new_width, new_height), Image.LANCZOS)
    else:
        starter_image_pil = generate_sd3_image(prompt)   


    # Move the image horizontally and vertically by 50%
    width, height = starter_image_pil.size
    horizontal_shift = width // 2
    vertical_shift = height // 2

    transformed_image_pil = ImageChops.offset(starter_image_pil, horizontal_shift, vertical_shift)

    # Create a new image with black background and white cross
    cross_image_pil = Image.new('RGB', (width, height), 'black')
    draw = ImageDraw.Draw(cross_image_pil)
    line_width = 75
    # Draw vertical line
    draw.rectangle([(width // 2 - line_width // 2, 0), (width // 2 + line_width // 2, height)], fill='white')
    # Draw horizontal line
    draw.rectangle([(0, height // 2 - line_width // 2), (width, height // 2 + line_width // 2)], fill='white')

    buffered = io.BytesIO()
    transformed_image_pil.save(buffered, format="JPEG")
    image_base64 = base64.b64encode(buffered.getvalue()).decode('utf-8')

    buffered = io.BytesIO()
    cross_image_pil.save(buffered, format="JPEG")
    cross_base64 = base64.b64encode(buffered.getvalue()).decode('utf-8')

    input = {
        "prompt": prompt + " smooth background",
        "negative_prompt": "worst quality, low quality, cartoons, sketch, ugly, lowres",
        "image": "data:image/jpeg;base64," + image_base64, 
        "mask": "data:image/jpeg;base64," + cross_base64, 
        "num_inference_steps": 25,
        "num_outputs": 3,
    }
    
    output = replicate.run(
        "lucataco/sdxl-inpainting:a5b13068cc81a89a4fbeefeccc774869fcb34df4dbc92c1555e0f2771d49dde7",
        input=input
    )
    
    images = []
    for i in range(min(len(output), 3)):
        image_url = output[i]
        response = requests.get(image_url)
        images.append(Image.open(io.BytesIO(response.content)))
    
    # Add empty images if fewer than 3 were returned
    while len(images) < 3:
        images.append(Image.new('RGB', (width, height), 'gray'))

    return images


demo = gr.Interface(fn=generate_pattern, inputs=["image", "text"], outputs=["image", "image", "image"])
demo.launch(share=False)