import gradio as gr
from diffusers import StableDiffusionPipeline
import torch

# Function to automatically switch between GPU and CPU
def load_model(model_id):
    if torch.cuda.is_available():
        device = "cuda"
        info = "Running on GPU (CUDA)"
    else:
        device = "cpu"
        info = "Running on CPU"
    
    # Load the model dynamically on the correct device
    pipe = StableDiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.float16 if device == "cuda" else torch.float32)
    pipe = pipe.to(device)
    
    return pipe, info

# Function for text-to-image generation with dynamic model ID and device info
def generate_image(model_id, prompt):
    pipe, info = load_model(model_id)
    image = pipe(prompt).images[0]
    return image, info

# Create the Gradio interface
with gr.Blocks() as demo:
    gr.Markdown("## Custom Text-to-Image Generator")

    with gr.Row():
        with gr.Column():
            model_id = gr.Textbox(label="Enter Model ID (e.g., nevreal/vMurderDrones)", placeholder="Model ID")
            prompt = gr.Textbox(label="Enter your prompt", placeholder="Describe the image you want to generate")
            generate_btn = gr.Button("Generate Image")
        
        with gr.Column():
            output_image = gr.Image(label="Generated Image")
            device_info = gr.Markdown()  # To display if GPU or CPU is used
    
    # Link the button to the image generation function
    generate_btn.click(fn=generate_image, inputs=[model_id, prompt], outputs=[output_image, device_info])

# Launch the app
demo.launch()