app.py

import os

import gradio as gr
import spaces
import torch
from diffusers import AutoPipelineForInpainting
from loguru import logger
from PIL import Image, ImageChops

SUPPORTED_MODELS = [
    "stabilityai/sdxl-turbo",
    "stabilityai/stable-diffusion-3-medium-diffusers",
    "stabilityai/stable-diffusion-xl-base-1.0",
    "stable-diffusion-v1-5/stable-diffusion-v1-5",
    "timbrooks/instruct-pix2pix",
]
DEFAULT_MODEL = "stabilityai/stable-diffusion-xl-base-1.0"


model = os.environ.get("MODEL_ID", DEFAULT_MODEL)
gpu_duration = int(os.environ.get("GPU_DURATION", 60))


def load_pipeline(model):
    return AutoPipelineForInpainting.from_pretrained(
        model, torch_dtype=torch.float16, use_safetensors=True, variant="fp16"
    )


logger.debug(f"Loading pipeline: {dict(model=model)}")
pipe = load_pipeline(model).to("cuda" if torch.cuda.is_available() else "mps")


@logger.catch(reraise=True)
@spaces.GPU(duration=gpu_duration)
def infer(
    prompt: str,
    image_editor: dict,
    negative_prompt: str,
    strength: float,
    num_inference_steps: int,
    guidance_scale: float,
    progress=gr.Progress(track_tqdm=True),
):
    logger.info(
        f"Starting image generation: {dict(model=model, prompt=prompt, image_editor=image_editor)}"
    )

    init_image: Image.Image = image_editor["background"].convert("RGB")

    # Downscale the image
    init_image.thumbnail((1024, 1024))

    mask_layer = image_editor["layers"][0]
    mask_image = Image.new("RGBA", mask_layer.size, "white")
    mask_image = Image.alpha_composite(mask_image, mask_layer).convert("RGB")
    mask_image = ImageChops.invert(mask_image)
    mask_image.thumbnail((1024, 1024))

    additional_args = {
        k: v
        for k, v in dict(
            strength=strength,
            num_inference_steps=num_inference_steps,
            guidance_scale=guidance_scale,
        ).items()
        if v
    }

    logger.debug(f"Generating image: {dict(prompt=prompt, **additional_args)}")

    images = pipe(
        prompt=prompt,
        image=init_image,
        mask_image=mask_image,
        negative_prompt=negative_prompt,
        **additional_args,
    ).images
    return images[0]


css = """
@media (max-width: 1280px) {
  #images-container {
    flex-direction: column;
  }
}
"""

with gr.Blocks(css=css) as demo:
    with gr.Column():
        gr.Markdown("# Inpainting")
        gr.Markdown(f"## Model: `{model}`")

        with gr.Row():
            prompt = gr.Text(
                label="Prompt",
                show_label=False,
                max_lines=1,
                placeholder="Enter your prompt",
                container=False,
            )

            run_button = gr.Button("Run", scale=0, variant="primary")

        with gr.Row(elem_id="images-container"):
            image_editor = gr.ImageMask(label="Initial image", type="pil")

            result = gr.Image(label="Result")

        with gr.Accordion("Advanced Settings", open=False):
            negative_prompt = gr.Text(
                label="Negative prompt",
                max_lines=1,
                placeholder="Enter a negative prompt",
            )

            with gr.Row():
                strength = gr.Slider(
                    label="Strength",
                    minimum=0.0,
                    maximum=1.0,
                    step=0.01,
                    value=0.0,
                )

                num_inference_steps = gr.Slider(
                    label="Number of inference steps",
                    minimum=0,
                    maximum=100,
                    step=1,
                    value=0,
                )

                guidance_scale = gr.Slider(
                    label="Guidance scale",
                    minimum=0.0,
                    maximum=100.0,
                    step=0.1,
                    value=0.0,
                )
    gr.on(
        triggers=[run_button.click, prompt.submit],
        fn=infer,
        inputs=[
            prompt,
            image_editor,
            negative_prompt,
            strength,
            num_inference_steps,
            guidance_scale,
        ],
        outputs=[result],
    )

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