Mac电脑手动安装原版Stable Diffusion，开启本地API调用生成图片

Mac 上安装 Stable Diffusion 的方法：
安装 Homebrew（如果尚未安装）
打开终端，执行以下命令：

/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

安装 Python 3.10.x

brew install python@3.10

安装 Git

brew install git

安装步骤
克隆 automatic1111 仓库

git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui.git
# 如果clone不下来，直接到github去下载再解压(unzip)也可以
cd stable-diffusion-webui

创建并激活虚拟环境

python3.10 -m venv venv
source venv/bin/activate

执行webui.sh

./webui.sh

等结束后会自动跳转127.0.0.1:7860

下载model

https://civitai.com/models

model放到stable diffusion webui/models/Stable Diffusion下

开启API

# 在webui-user.sh添加
# Stable Diffusion WebUI启动脚本
export COMMANDLINE_ARGS="--lowvram --precision full --no-half --skip-torch-cuda-test --disable-nan-check --api"

再启动./webui.sh

编写本地调用api代码

import requests
import json
import base64
from PIL import Image
from io import BytesIO
import os

# API地址（默认端口7860）
API_URL = "http://127.0.0.1:7860"

# 1. 文本生成图像（txt2img）
def txt2img(prompt, negative_prompt="", steps=20, width=512, height=512, sampler="Euler a"):
    payload = {
        "prompt": prompt,
        "negative_prompt": negative_prompt,
        "steps": steps,
        "width": width,
        "height": height,
        "sampler_name": sampler,
        "cfg_scale": 7,  # 提示词相关性，值越高越贴近提示
        "seed": -1,  # -1表示随机种子
    }

    response = requests.post(f"{API_URL}/sdapi/v1/txt2img", json=payload)
    if response.status_code == 200:
        result = response.json()
        for i, img_data in enumerate(result["images"]):
            image = Image.open(BytesIO(base64.b64decode(img_data)))
            save_path = f"txt2img_result_{i}.png"
            image.save(save_path)
            print(f"图片已保存至: {save_path}")
        return result
    else:
        print(f"请求失败，状态码: {response.status_code}")
        print(response.text)
        return None

# 2. 图像生成图像（img2img）
def img2img(prompt, image_path, negative_prompt="", steps=20, strength=0.7, sampler="Euler a"):
    # 读取图像并转换为base64
    with open(image_path, "rb") as f:
        image_data = f.read()
    image_base64 = base64.b64encode(image_data).decode('utf-8')

    payload = {
        "prompt": prompt,
        "negative_prompt": negative_prompt,
        "init_images": [image_base64],
        "steps": steps,
        "sampler_name": sampler,
        "strength": strength,  # 0.1-0.9，数值越高与原图差异越大
        "cfg_scale": 7,
        "seed": -1,
    }

    response = requests.post(f"{API_URL}/sdapi/v1/img2img", json=payload)
    if response.status_code == 200:
        result = response.json()
        for i, img_data in enumerate(result["images"]):
            image = Image.open(BytesIO(base64.b64decode(img_data)))
            save_path = f"img2img_result_{i}.png"
            image.save(save_path)
            print(f"图片已保存至: {save_path}")
        return result
    else:
        print(f"请求失败，状态码: {response.status_code}")
        print(response.text)
        return None

# 3. 获取可用模型列表
def get_models():
    response = requests.get(f"{API_URL}/sdapi/v1/sd-models")
    if response.status_code == 200:
        return response.json()
    return []

# 4. 切换模型
def switch_model(model_name):
    payload = {"sd_model_checkpoint": model_name}
    response = requests.post(f"{API_URL}/sdapi/v1/options", json=payload)
    return response.status_code == 200

# 示例调用
if __name__ == "__main__":
    # 示例1: 文本生成图像
    txt2img(
        prompt="a beautiful landscape, mountains, lake, detailed, realistic, 4k",
        negative_prompt="blurry, low quality, deformed",
        steps=30,
        width=768,
        height=512
    )

    # 示例2: 图像生成图像（取消注释并提供图片路径）
    # img2img(
    #     prompt="make it more vibrant, add birds in the sky",
    #     image_path="input_image.png",
    #     strength=0.6
    # )

    # 示例3: 获取可用模型
    # models = get_models()
    # print("可用模型:", [model["title"] for model in models])

    # 示例4: 切换模型
    # switch_model("v1-5-pruned-emaonly.ckpt")

运行后

这样就可以编写本地脚本，实现自动化生图了，而无需在webui上一张张图自己生成了。