app.py

# import torch
# from diffusers import StableDiffusionPipeline, ControlNetModel, StableDiffusionControlNetPipeline
# from PIL import Image
# import gradio as gr

# # 自动选择设备
# device = "cuda" if torch.cuda.is_available() else "cpu"
# print(f"Using device: {device}")

# # 加载 ControlNet 模型
# controlnet = ControlNetModel.from_pretrained(
#     "lllyasviel/sd-controlnet-canny", torch_dtype=torch.float32
# )

# # 加载 Stable Diffusion + ControlNet
# pipe = StableDiffusionControlNetPipeline.from_pretrained(
#     "CompVis/stable-diffusion-v1-4",
#     controlnet=controlnet,
#     torch_dtype=torch.float32
# ).to(device)

# # CPU 下节省显存
# pipe.enable_attention_slicing()

# # 文生图生成函数
# def generate_image(prompt, num_steps=20, height=256, width=256):
#     """
#     prompt: str, 文本描述
#     num_steps: int, 推理步数（CPU 可少一些）
#     height, width: int, 输出图像分辨率
#     """
#     image = pipe(prompt, num_inference_steps=num_steps, height=height, width=width).images[0]
#     return image

# # Gradio 界面
# interface = gr.Interface(
#     fn=generate_image,
#     inputs=[
#         gr.Textbox(label="Prompt", placeholder="Enter text prompt here..."),
#         gr.Slider(5, 50, value=20, step=1, label="Inference Steps"),
#         gr.Slider(128, 512, value=256, step=64, label="Height"),
#         gr.Slider(128, 512, value=256, step=64, label="Width"),
#     ],
#     outputs=gr.Image(type="pil"),
#     title="Text2Image Demo (v1-4 + ControlNet)",
#     description="Generate images from text using Stable Diffusion v1-4 + ControlNet (CPU/GPU compatible)"
# )

# interface.launch()

import torch
from diffusers import StableDiffusionPipeline, ControlNetModel, StableDiffusionControlNetPipeline
from PIL import Image
import gradio as gr
import numpy as np

# 自动选择设备
device = "cuda" if torch.cuda.is_available() else "cpu"
print(f"Using device: {device}")

# 加载 ControlNet 模型
controlnet = ControlNetModel.from_pretrained(
    "lllyasviel/sd-controlnet-canny", torch_dtype=torch.float32
)

# 加载 Stable Diffusion + ControlNet
pipe = StableDiffusionControlNetPipeline.from_pretrained(
    "CompVis/stable-diffusion-v1-4",
    controlnet=controlnet,
    torch_dtype=torch.float32
).to(device)

# CPU 下节省显存
pipe.enable_attention_slicing()

# 文生图生成函数
def generate_image(prompt, num_steps=20, height=256, width=256):
    dummy_image = Image.fromarray(np.zeros((height, width, 3), dtype=np.uint8))
    image = pipe(prompt, num_inference_steps=num_steps, height=height, width=width,image=dummy_image).images[0]
    return image

# Gradio 界面
interface = gr.Interface(
    fn=generate_image,
    inputs=[
        gr.Textbox(label="Prompt", placeholder="Enter text prompt here..."),
        gr.Slider(5, 50, value=20, step=1, label="Inference Steps"),
        gr.Slider(128, 512, value=256, step=64, label="Height"),
        gr.Slider(128, 512, value=256, step=64, label="Width"),
    ],
    outputs=gr.Image(type="pil"),
    title="Text2Image Demo (v1-4 + ControlNet)",
    description="Generate images from text using Stable Diffusion v1-4 + ControlNet"
)

interface.launch()