# Create a text-to-image interface with options to choose schedulers, sampling steps, and no default dark theme.
import gradio as gr
import numpy as np
from PIL import Image
import time

# Define a function to generate an image based on the input text, scheduler, and sampling steps.
def generate_image(text, scheduler, sampling_steps):
    # For demonstration purposes, we will generate a random image.
    # In a real application, this function would call a text-to-image model.
    rng = np.random.default_rng()
    image = rng.random(size=(600, 600, 3))
    
    # Simulate the effect of different schedulers and sampling steps
    if scheduler == "Euler a":
        image = np.clip(image * (1 + sampling_steps / 100), 0, 1)
    elif scheduler == "DPM++ 2M Karras":
        image = np.clip(image * (1 + sampling_steps / 50), 0, 1)
    elif scheduler == "Euler":
        image = np.clip(image * (1 + sampling_steps / 20), 0, 1)
    
    # Generate a unique filename for the image using the current timestamp
    timestamp = int(time.time() * 1000)
    filename = f"image_{timestamp}.webp"
    
    # Convert the numpy array to a PIL image and save it in WEBP format
    image_pil = Image.fromarray((image * 255).astype(np.uint8))
    image_pil.save(filename, format="WEBP")
    
    return filename

# Create a Gradio interface with text input, scheduler dropdown, sampling steps slider, and no default dark theme.
with gr.Blocks() as demo:
    with gr.Row():
        text_input = gr.Textbox(label="Input Text")
        scheduler_dropdown = gr.Dropdown(choices=["Euler a", "DPM++ 2M Karras", "Euler"], value="Euler a", label="Scheduler")
        sampling_steps_slider = gr.Slider(1, 100, value=50, step=1, label="Sampling Steps")
    
    with gr.Row():
        generate_button = gr.Button("Generate Image")
        output_image = gr.Image(label="Generated Image")
    
    # Define the event listener for the generate button.
    generate_button.click(
        fn=generate_image,
        inputs=[text_input, scheduler_dropdown, sampling_steps_slider],
        outputs=output_image
    )

# Launch the Gradio interface.
if __name__ == "__main__":
    gr.load("models/stabilityai/stable-diffusion-3.5-large").launch()
    demo.launch(show_error=True)