添加macos打包时候需要进行修改的文件

Author: YaoFANGUK

backend/config.py

@@ -1,7 +1,8 @@
 
 import os
+import sys
 from pathlib import Path
-from qfluentwidgets import (qconfig, ConfigItem, QConfig, OptionsValidator, BoolValidator, OptionsConfigItem, 
+from qfluentwidgets import (qconfig, ConfigItem, QConfig, OptionsValidator, BoolValidator, OptionsConfigItem,
                             EnumSerializer, RangeValidator, RangeConfigItem, ConfigValidator)
 from backend.tools.constant import SubtitleArea, VideoSubFinderDecoder
 import configparser
@@ -14,14 +15,20 @@
 PROJECT_UPDATE_URLS = [
     "https://api.github.com/repos/YaoFANGUK/video-subtitle-extractor/releases/latest",
     "https://accelerate.xdow.net/api/repos/YaoFANGUK/video-subtitle-extractor/releases/latest",
-] 
+]
 # 硬件加速选项开关
 HARDWARD_ACCELERATION_OPTION = True
 
+# 项目的base目录（打包后资源在 sys._MEIPASS/backend 下）
+if getattr(sys, 'frozen', False):
+    BASE_DIR = os.path.join(sys._MEIPASS, 'backend')
+else:
+    BASE_DIR = str(Path(os.path.abspath(__file__)).parent)
+
 # 读取界面语言配置
 tr = configparser.ConfigParser()
 
-TRANSLATION_FILE = os.path.join(os.path.dirname(os.path.abspath(__file__)), 'interface', f"en.ini")
+TRANSLATION_FILE = os.path.join(BASE_DIR, 'interface', f"en.ini")
 tr.read(TRANSLATION_FILE, encoding='utf-8')
 
 class Config(QConfig):
@@ -97,17 +104,20 @@ class Config(QConfig):
     # VideoSubFinder 视频解码组件
     videoSubFinderDecoder = OptionsConfigItem("Main", "VideoSubFinderDecoder", VideoSubFinderDecoder.OPENCV, OptionsValidator(VideoSubFinderDecoder), EnumSerializer(VideoSubFinderDecoder))
 
-CONFIG_FILE = 'config/config.json'
+# 打包后 app bundle 内为只读，配置文件需存到用户可写目录
+if getattr(sys, 'frozen', False):
+    _config_dir = os.path.join(os.path.expanduser('~'), 'Library', 'Application Support', 'VideoSubtitleExtractor')
+    os.makedirs(_config_dir, exist_ok=True)
+    CONFIG_FILE = os.path.join(_config_dir, 'config.json')
+else:
+    CONFIG_FILE = 'config/config.json'
 config = Config()
 qconfig.load(CONFIG_FILE, config)
 
 # 读取界面语言配置
 tr = configparser.ConfigParser()
 
-TRANSLATION_FILE = os.path.join(os.path.dirname(os.path.abspath(__file__)), 'interface', f"{config.interface.value}.ini")
+TRANSLATION_FILE = os.path.join(BASE_DIR, 'interface', f"{config.interface.value}.ini")
 tr.read(TRANSLATION_FILE, encoding='utf-8')
 
-# 项目的base目录
-BASE_DIR = str(Path(os.path.abspath(__file__)).parent)
-
 os.environ['KMP_DUPLICATE_LIB_OK'] = 'True'

backend/main.py

@@ -26,6 +26,7 @@
 from backend.config import *
 from backend.tools.hardware_accelerator import HardwareAccelerator
 from backend.tools import reformat
+
 from backend.tools.ocr import OcrRecogniser, get_coordinates
 from backend.tools import subtitle_ocr
 from backend.tools.paddle_model_config import PaddleModelConfig
@@ -38,6 +39,7 @@
 import time
 import pysrt
 
+
 class SubtitleExtractor:
     """
     视频字幕提取类
@@ -110,16 +112,18 @@ def run(self):
         self.update_progress(ocr=0, frame_extract=0)
         self.append_output('-----------------------------')
         # 打印识别语言与识别模式
-        self.append_output(f"  {tr['Main']['RecSubLang']}：{config.language.value}  |  {tr['Main']['RecMode']}：{config.mode.value}")
+        self.append_output(
+            f"  {tr['Main']['RecSubLang']}：{config.language.value}  |  {tr['Main']['RecMode']}：{config.mode.value}")
         # 如果使用GPU加速，则打印GPU加速提示
         if self.hardware_accelerator.has_accelerator():
             self.append_output(f"  {tr['Main']['AcceleratorON'].format(self.hardware_accelerator.accelerator_name)}")
 
         # 打印视频帧数与帧率
         self.append_output(f"  {tr['Main']['FrameCount']}：{self.frame_count}"
-              f"  |  {tr['Main']['FrameRate']}：{self.fps}")
+                           f"  |  {tr['Main']['FrameRate']}：{self.fps}")
         # 打印加载模型信息
-        self.append_output(f"  DET: {os.path.basename(self.model_config.DET_MODEL_PATH)}  |  REC: {os.path.basename(self.model_config.REC_MODEL_PATH)}")
+        self.append_output(
+            f"  DET: {os.path.basename(self.model_config.DET_MODEL_PATH)}  |  REC: {os.path.basename(self.model_config.REC_MODEL_PATH)}")
         self.append_output('-----------------------------')
         # 打印视频帧提取开始提示
         self.append_output(tr['Main']['StartProcessFrame'])
@@ -197,31 +201,31 @@ def capture_frame_with_subtitle_area(self):
         # 确保输出目录存在
         if not os.path.exists(self.temp_output_dir):
             os.makedirs(self.temp_output_dir)
-            
+
         # 确保视频已打开
         if not self.video_cap.isOpened():
             self.video_cap = cv2.VideoCapture(self.video_path)
-            
+
         # 将视频指针设置到第一帧
         # self.video_cap.set(cv2.CAP_PROP_POS_FRAMES, 0)
-        
+
         # 读取第一帧
         ret, frame = self.video_cap.read()
-        
+
         if ret:
             # 如果有字幕区域，绘制矩形
             sub_area = self.sub_area
             if sub_area is not None:
                 # 绘制绿色矩形框
                 cv2.rectangle(frame, (sub_area.xmin, sub_area.ymin), (sub_area.xmax, sub_area.ymax), (0, 255, 0), 2)
                 # 添加文字标注
-                cv2.putText(frame, "Subtitle Area", (sub_area.xmin, sub_area.ymin - 10), 
-                           cv2.FONT_HERSHEY_SIMPLEX, 0.9, (0, 255, 0), 2)
-            
+                cv2.putText(frame, "Subtitle Area", (sub_area.xmin, sub_area.ymin - 10),
+                            cv2.FONT_HERSHEY_SIMPLEX, 0.9, (0, 255, 0), 2)
+
             # 保存图像
             output_path = os.path.join(self.temp_output_dir, 'sub_area.jpg')
             cv2.imwrite(output_path, frame)
-            
+
             # 重置视频指针到第一帧
             self.video_cap.set(cv2.CAP_PROP_POS_FRAMES, 0)
 
@@ -308,7 +312,8 @@ def extract_frame_by_det(self):
                     dt_box, rec_res = self.ocr.predict(frame)
                     area_text1 = "".join(self.__get_area_text((dt_box, rec_res)))
                     if start_frame_no not in compare_ocr_result_cache.keys():
-                        compare_ocr_result_cache[current_frame_no] = {'text': area_text1, 'dt_box': dt_box, 'rec_res': rec_res}
+                        compare_ocr_result_cache[current_frame_no] = {'text': area_text1, 'dt_box': dt_box,
+                                                                      'rec_res': rec_res}
                         frame_lru_list.append((frame, current_frame_no))
                         ocr_args_list.append((self.frame_count, current_frame_no))
                         # 缓存头帧
@@ -327,7 +332,8 @@ def extract_frame_by_det(self):
                 # 如果在找结束帧的时候
                 if is_finding_end_frame_no:
                     # 判断该帧与头帧ocr内容是否一致,若不一致则找到尾，尾巴为前一帧
-                    if not self._compare_ocr_result(compare_ocr_result_cache, None, start_frame_no, frame, current_frame_no):
+                    if not self._compare_ocr_result(compare_ocr_result_cache, None, start_frame_no, frame,
+                                                    current_frame_no):
                         is_finding_end_frame_no = False
                         is_finding_start_frame_no = True
                         end_frame_no = current_frame_no - 1
@@ -349,7 +355,7 @@ def extract_frame_by_det(self):
                 frame_lru_list.pop(0)
 
             # if len(start_end_frame_no) > 0:
-                # self.append_output(start_end_frame_no)
+            # self.append_output(start_end_frame_no)
 
             while len(ocr_args_list) > 1:
                 total_frame_count, ocr_info_frame_no = ocr_args_list.pop(0)
@@ -384,6 +390,7 @@ def extract_frame_by_vsf(self):
         if self.video_cap:
             self.video_cap.release()
             self.video_cap = None
+
         def count_process():
             duration_ms = (self.frame_count / self.fps) * 1000
             last_total_ms = 0
@@ -481,10 +488,11 @@ def vsf_output(out, ):
             # 计算进度
             try:
                 self.vsf_running = True
-                Thread(target=count_process, daemon=True).start()       
+                Thread(target=count_process, daemon=True).start()
                 # 已知BUG: test_chinese_cht.flv在net drive上会导致无法停止, 但在本地不会, 可能是vsf的原因
                 p = subprocess.Popen(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE, bufsize=1,
-                                    close_fds='posix' in sys.builtin_module_names, shell=False, creationflags=subprocess.CREATE_NEW_PROCESS_GROUP)
+                                     close_fds='posix' in sys.builtin_module_names, shell=False,
+                                     creationflags=subprocess.CREATE_NEW_PROCESS_GROUP)
                 ProcessManager.instance().add_process(p)
                 self.manage_process(p.pid)
                 p.wait()
@@ -500,14 +508,15 @@ def vsf_output(out, ):
             self.vsf_running = True
             try:
                 p = subprocess.Popen(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE, bufsize=1,
-                                    close_fds='posix' in sys.builtin_module_names, shell=True,
-                                    start_new_session=True)
+                                     close_fds='posix' in sys.builtin_module_names, shell=True,
+                                     start_new_session=True)
                 Thread(target=vsf_output, daemon=True, args=(p.stderr,)).start()
                 ProcessManager.instance().add_process(p)
                 self.manage_process(p.pid)
                 p.wait()
             finally:
                 self.vsf_running = False
+
     def filter_watermark(self):
         """
         去除原始字幕文本中的水印区域的文本
@@ -773,7 +782,9 @@ def _remove_duplicate_subtitle(self):
             while idx_j < content_list_len:
                 # 计算当前行与下一行的Levenshtein距离
                 # 判决idx_j的下一帧是否与idx_i不同，若不同（或者是最后一帧）则找到结束帧
-                if idx_j + 1 == content_list_len or ratio(i.content.replace(' ', ''), content_list[idx_j + 1].content.replace(' ', '')) < (config.thresholdTextSimilarity.value / 100.0):
+                if idx_j + 1 == content_list_len or ratio(i.content.replace(' ', ''),
+                                                          content_list[idx_j + 1].content.replace(' ', '')) < (
+                        config.thresholdTextSimilarity.value / 100.0):
                     # 若找到终点帧,定义字幕结束帧帧号
                     end_frame = content_list[idx_j].no
                     if not self.use_vsf:
@@ -836,18 +847,21 @@ def _unite_coordinates(self, coordinates_list):
         indexed = sorted(enumerate(coordinates_list), key=lambda x: x[1][0])
         # parent数组用于并查集
         parent = list(range(n))
+
         def find(i):
             while parent[i] != i:
                 parent[i] = parent[parent[i]]
                 i = parent[i]
             return i
+
         def union(i, j):
             ri, rj = find(i), find(j)
             if ri != rj:
                 # 保留较小索引的坐标作为代表
                 if ri > rj:
                     ri, rj = rj, ri
                 parent[rj] = ri
+
         # 滑动窗口：xmin已排序，只要xmin差值超过容忍度就移动左边界
         left = 0
         for right in range(n):
@@ -1001,16 +1015,18 @@ def get_ocr_progress():
                 # self.append_output(f'recv total_ms:{total_ms}')
                 if current_frame_no == -1:
                     return
+
         options = {
             'REC_CHAR_TYPE': config.language.value,
             'DROP_SCORE': config.dropScore.value / 100.0,
             'SUB_AREA_DEVIATION_RATE': config.subtitleAreaDeviationRate.value / 100.0,
             'DEBUG_OCR_LOSS': config.debugOcrLoss.value,
             'HARDWARD_ACCELERATOR': self.hardware_accelerator,
         }
-        process, task_queue, progress_queue = subtitle_ocr.async_start(self.video_path, self.raw_subtitle_path, self.sub_area, options)
+        process, task_queue, progress_queue = subtitle_ocr.async_start(self.video_path, self.raw_subtitle_path,
+                                                                       self.sub_area, options)
         ProcessManager.instance().add_process(process)
-        self.manage_process(process.pid)
+        self.manage_process(getattr(process, 'pid', None))
         self.subtitle_ocr_task_queue = task_queue
         self.subtitle_ocr_progress_queue = progress_queue
         # 开启线程负责更新OCR进度
@@ -1035,23 +1051,23 @@ def append_output(self, *args):
     def add_progress_listener(self, listener):
         """
         添加进度监听器
-        
+
         Args:
             listener: 一个回调函数，接收参数 (progress_ocr, progress_frame_extract, progress_total, isFinished)
         """
         if listener not in self.progress_listeners:
             self.progress_listeners.append(listener)
-    
+
     def remove_progress_listener(self, listener):
         """
         移除进度监听器
-        
+
         Args:
             listener: 要移除的监听器函数
         """
         if listener in self.progress_listeners:
             self.progress_listeners.remove(listener)
-            
+
     def notify_progress_listeners(self):
         """
         通知所有进度监听器当前进度
@@ -1065,6 +1081,7 @@ def notify_progress_listeners(self):
     def manage_process(pid):
         pass
 
+
 if __name__ == '__main__':
     multiprocessing.set_start_method("spawn")
     # 提示用户输入视频路径

backend/tools/ocr.py

@@ -1,10 +1,34 @@
 import os
+import traceback
 from backend.config import *
 import importlib
 from paddleocr import PaddleOCR
 from backend.tools.hardware_accelerator import HardwareAccelerator
 from backend.tools.paddle_model_config import PaddleModelConfig
 
+# PyInstaller compatibility: bypass paddlex dependency checks
+# importlib.metadata can't find dist-info dirs in the bundle, causing false alarms.
+# All deps are bundled, so the checks are safe to skip.
+try:
+    import paddlex.utils.deps as _pdeps
+    import importlib.util as _il_util
+
+    # Patch is_dep_available: use find_spec for special deps, True for everything else
+    def _patched_is_dep_available(dep, /, check_version=False):
+        _special = {"paddlepaddle": "paddle", "paddle-custom-device": "paddle_custom_device",
+                    "ultra-infer": "ultra_infer", "fastdeploy": "fastdeploy",
+                    "onnxruntime": "onnxruntime"}
+        if dep in _special:
+            return _il_util.find_spec(_special[dep]) is not None
+        return True
+
+    _pdeps.is_dep_available = _patched_is_dep_available
+    # Safety net: also bypass require_extra and require_deps
+    _pdeps.require_extra = lambda *a, **kw: None
+    _pdeps.require_deps = lambda *a, **kw: None
+except Exception as e:
+    print(f"Warning: failed to patch paddlex deps: {e}")
+
 # 加载文本检测+识别模型
 class OcrRecogniser:
     def __init__(self):
@@ -105,7 +129,15 @@ def init_model(self):
         if model_config.REC_MODEL_NAME:
             kwargs['text_recognition_model_name'] = model_config.REC_MODEL_NAME
 
-        return PaddleOCR(**kwargs)
+        try:
+            return PaddleOCR(**kwargs)
+        except Exception as e:
+            # Print full error chain for debugging
+            print(f"Error initializing PaddleOCR: {e}")
+            if e.__cause__:
+                print(f"  Caused by: {e.__cause__}")
+            traceback.print_exc()
+            raise
 
 
 def get_coordinates(dt_box):

backend/tools/path_utils.py

@@ -0,0 +1,25 @@
+import os
+import sys
+
+
+def resource_path(relative_path):
+    """Get absolute path to a bundled resource.
+
+    When running in a PyInstaller bundle, resources are extracted to sys._MEIPASS.
+    In development, resolve relative to the project root (parent of backend/).
+    """
+    if hasattr(sys, '_MEIPASS'):
+        return os.path.join(sys._MEIPASS, relative_path)
+    # dev mode: project root is two levels up from this file
+    return os.path.join(os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__)))), relative_path)
+
+
+def app_path(relative_path):
+    """Get path relative to the executable directory for user-writable data.
+
+    In a PyInstaller bundle, this is next to the .app/.exe.
+    In development, same as resource_path (project root).
+    """
+    if getattr(sys, 'frozen', False):
+        return os.path.join(os.path.dirname(sys.executable), relative_path)
+    return resource_path(relative_path)