import gradio as gr
import torch
from diffusers import DiffusionPipeline

device = "cpu"
if torch.cuda.is_available():
    device = "cuda"

model_id = "SimianLuo/LCM_Dreamshaper_v7"

pipe = DiffusionPipeline.from_pretrained(model_id)
pipe.to(device)

def generate_image(prompt, width, height, steps):
    result = pipe(
        prompt=prompt,
        width=int(width),
        height=int(height),
        num_inference_steps=int(steps),
        guidance_scale=8.0,
        lcm_origin_steps=50,
        output_type="pil"
    ).images[0]
    return result

with gr.Blocks() as demo:
    gr.Markdown("## LCM Image Generator (CPU Optimized)")
    
    with gr.Row():
        with gr.Column():
            prompt_input = gr.Textbox(label="Prompt", placeholder="A futuristic city with neon lights...")
            
            with gr.Row():
                width_slider = gr.Slider(minimum=256, maximum=768, step=64, value=512, label="Width")
                height_slider = gr.Slider(minimum=256, maximum=768, step=64, value=512, label="Height")
            
            steps_slider = gr.Slider(minimum=1, maximum=15, step=1, value=4, label="Inference Steps (Quality)")
            
            generate_btn = gr.Button("Generate Image")
        
        with gr.Column():
            image_output = gr.Image(label="Generated Image")

    generate_btn.click(
        fn=generate_image,
        inputs=[prompt_input, width_slider, height_slider, steps_slider],
        outputs=image_output
    )

demo.launch()