model_id = "stabilityai/stable-diffusion-xl-base-1.0"
scheduler = EulerAncestralDiscreteScheduler.from_pretrained(model_id, subfolder="scheduler")
pipe = StableDiffusionXLPipeline.from_pretrained(model_id, scheduler=scheduler, torch_dtype=torch.float16, revision="fp16")
pipe = pipe.to("cuda")

def generate_image(prompt: str, negative_prompt: str = "", height: int = 512, width: int = 512, num_inference_steps: int = 50, guidance_scale: float = 7.5, num_images_per_prompt: int = 1) -> Tuple[Image.Image, str]:
    generator = torch.Generator(device="cuda").manual_seed(random.randint(0, 2**32 - 1))
    image = pipe(
        prompt=prompt,
        negative_prompt=negative_prompt,
        height=height,
        width=width,
        num_inference_steps=num_inference_steps,
        guidance_scale=guidance_scale,
        num_images_per_prompt=num_images_per_prompt,
        generator=generator,
    ).images[0]

    image_id = str(uuid.uuid4())
    image_path = f"/tmp/{image_id}.png"
    image.save(image_path)

    return image, image_path

def gradio_interface():
    with gr.Blocks(css=css) as demo:
        gr.Markdown("## Gere imagens usando Stable Diffusion XL")
        with gr.Row():
            with gr.Column():
                prompt = gr.Textbox(label="Prompt", placeholder="Digite o prompt aqui...")
                negative_prompt = gr.Textbox(label="Negative Prompt", placeholder="Digite o negative prompt aqui...")
                run_button = gr.Button("Gerar Imagem")
            with gr.Column():
                result = gr.Image(label="Imagem Gerada")

        run_button.click(
            fn=lambda p, np: generate_image(p, np)[0],
            inputs=[prompt, negative_prompt],
            outputs=result,
        )

    return demo

if __name__ == "__main__":
    demo = gradio_interface()
    demo.launch()