app.py

import os
from flask import Flask, request, send_file
from io import BytesIO
from PIL import Image, ImageChops
from diffusers import StableDiffusionPipeline, EulerDiscreteScheduler
import torch

# Flaskアプリケーションの初期化
app = Flask(__name__)

# グローバルなモデル辞書
models_load = {}

def load_model(model_name):
    global models_load
    if model_name not in models_load:
        try:
            scheduler = EulerDiscreteScheduler.from_pretrained(model_name, subfolder="scheduler")
            pipe = StableDiffusionPipeline.from_pretrained(model_name, scheduler=scheduler, torch_dtype=torch.float16)
            pipe = pipe.to("cuda")
            models_load[model_name] = pipe
            print(f"Model {model_name} loaded successfully.")
        except Exception as error:
            print(f"Error loading model {model_name}: {error}")
            models_load[model_name] = None

def gen_fn(model_str, prompt, negative_prompt=None, noise=None, cfg_scale=None, num_inference_steps=None):
    if model_str not in models_load:
        load_model(model_str)

    if model_str in models_load and models_load[model_str] is not None:
        if noise == "random":
            noise = str(randint(0, 99999999999))
        full_prompt = f'{prompt} {noise}' if noise else prompt

        print(f"Prompt: {full_prompt}")

        try:
            # モデル呼び出し
            result = models_load[model_str](full_prompt, num_inference_steps=num_inference_steps)
            image = result.images[0]  # 生成された画像を取得

            # Check if the image is completely black
            black = Image.new('RGB', image.size, (0, 0, 0))
            if ImageChops.difference(image, black).getbbox() is None:
                return None, 'The image is completely black.'

            return image, None

        except Exception as e:
            print("Error generating image:", e)
            return None, f"Error generating image: {e}"
    else:
        print(f"Model {model_str} not found")
        return None, f"Model {model_str} not found"

@app.route('/', methods=['GET'])
def home():
    prompt = request.args.get('prompt', '')
    model = request.args.get('model', '')
    noise = request.args.get('noise', None)
    num_inference_steps = request.args.get('steps', 50)  # デフォルト値を設定

    try:
        num_inference_steps = int(num_inference_steps)
    except ValueError:
        return 'Invalid "steps" parameter. It should be an integer.', 400
    
    if not model:
        return 'Please provide a "model" query parameter in the URL.', 400

    if not prompt:
        return 'Please provide a "prompt" query parameter in the URL.', 400
    
    # Generate the image
    image, error_message = gen_fn(model, prompt, noise=noise, num_inference_steps=num_inference_steps)
    if error_message:
        return error_message, 400
    
    if isinstance(image, Image.Image):
        img_io = BytesIO()
        image.save(img_io, format='PNG')
        img_io.seek(0)
        return send_file(img_io, mimetype='image/png', as_attachment=False)
    
    return 'Failed to generate image.', 500

if __name__ == '__main__':
    app.run(host='0.0.0.0', port=7860)