diff --git a/cuda-hasten.py b/cuda-hasten.py
new file mode 100644
index 0000000..c2c0923
--- /dev/null
+++ b/cuda-hasten.py
@@ -0,0 +1,45 @@
+# 使用CUDA进行加速，但实际效果是减速 ：（
+
+import os
+import torch
+from torchvision.io.image import read_image, write_png
+from torchvision.transforms.functional import to_pil_image
+from PIL import Image
+
+# 设置允许的最大像素数，例如设置为1亿像素
+Image.MAX_IMAGE_PIXELS = 1000000000
+def convert_jpg_to_webp_cuda(input_folder, output_folder):
+    # 确保输出文件夹存在
+    os.makedirs(output_folder, exist_ok=True)
+    
+    # 遍历输入文件夹中的所有文件
+    for filename in os.listdir(input_folder):
+        # 检查文件是否为jpg或JPEG格式
+        if filename.endswith(".jpg") or filename.endswith(".jpeg"):
+            # 构建完整的文件路径
+            input_path = os.path.join(input_folder, filename)
+            # 获取新的文件名，将.jpg/.jpeg替换为.webp
+            webp_filename = filename.rsplit('.', 1)[0] + '.webp'
+            output_path = os.path.join(output_folder, webp_filename)
+            
+            try:
+                # 使用PyTorch读取图像，尝试使用CUDA
+                device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+                image_tensor = read_image(input_path).to(device)
+                
+                # 注意：这里我们先转换为PIL Image，因为torchvision目前不直接支持WEBP输出
+                # 实际上，这一步会将处理移回CPU，但整个流程仍可能受益于CUDA加速的预处理
+                pil_image = to_pil_image(image_tensor.cpu())
+                
+                # 使用Pillow保存为WEBP
+                pil_image.save(output_path, format="WEBP")
+                print(f"Converted {filename} to {webp_filename} using CUDA")
+            except Exception as e:
+                print(f"Cannot convert {filename}: {e}")
+
+# 指定输入和输出文件夹
+input_folder = 'input'
+output_folder = 'output_cuda'
+
+# 调用函数进行转换
+convert_jpg_to_webp_cuda(input_folder, output_folder)
\ No newline at end of file
diff --git a/main.py b/main.py
new file mode 100644
index 0000000..77ac84d
--- /dev/null
+++ b/main.py
@@ -0,0 +1,56 @@
+from PIL import Image
+import os
+import concurrent.futures
+from PIL import Image
+
+# 设置允许的最大像素数，例如设置为1亿像素
+Image.MAX_IMAGE_PIXELS = 1000000000
+def validate_folder_path(folder_path):
+    """
+    验证文件夹路径的安全性，避免路径遍历攻击。
+    """
+    clean_path = os.path.abspath(os.path.expanduser(folder_path))
+    if not clean_path.startswith(os.path.abspath(os.getcwd())):
+        raise ValueError("不安全的路径访问尝试。")
+    return clean_path
+
+def convert_image_to_webp(input_path, output_path):
+    """
+    单个图像转换函数，为了异常处理和代码清晰。
+    """
+    try:
+        with Image.open(input_path) as img:
+            img.save(output_path, "WEBP")
+        print(f"Converted {os.path.basename(input_path)} to {os.path.basename(output_path)}")
+    except Exception as e:  # 捕获更广泛的异常
+        print(f"Cannot convert {os.path.basename(input_path)}: {e}")
+
+def convert_jpg_to_webp(input_folder, output_folder, max_workers=None):
+    input_folder = validate_folder_path(input_folder)
+    output_folder = validate_folder_path(output_folder)
+
+    if not os.path.exists(output_folder):
+        try:
+            os.makedirs(output_folder)
+        except Exception as e:
+            print(f"无法创建输出文件夹 {output_folder}: {e}")
+            return
+
+    # 使用集合来简化文件格式检查
+    valid_extensions = {"jpg", "jpeg", "png" , "JPG"}
+
+    # 使用并发执行来提高性能
+    with concurrent.futures.ThreadPoolExecutor(max_workers=max_workers) as executor:
+        for filename in os.listdir(input_folder):
+            if filename.endswith(tuple(valid_extensions)):
+                input_path = os.path.join(input_folder, filename)
+                webp_filename = os.path.splitext(filename)[0] + '.webp'
+                output_path = os.path.join(output_folder, webp_filename)
+                executor.submit(convert_image_to_webp, input_path, output_path)
+
+# 指定输入和输出文件夹
+input_folder = 'input'
+output_folder = 'output'
+
+	# 调用函数进行转换，使用6个线程并发转换（根据实际情况调整）
+convert_jpg_to_webp(input_folder, output_folder, max_workers=6)
\ No newline at end of file