diff --git a/README.md b/README.md
index 38acc2f..b971b2f 100755
--- a/README.md
+++ b/README.md
@@ -201,7 +201,7 @@ cd <源码所在目录>
 ##### (4) macOS 运行 (Apple Silicon)
 - 适用于 macOS (Apple Silicon) 设备
 - macOS (Intel) 请使用CPU, 强行使用GPU只会更慢
-- macOS (Apple Silicon)上字幕检测精准模式的模型(ch_det_fast/model.onnx)精度似乎不太理想, 推荐使用快速模式
+- macOS (Apple Silicon)上字幕检测PP-OCRv4-Server模型精度似乎不太理想, 推荐使用其他模型
   ```shell
   pip install paddlepaddle==3.0.0 -i https://www.paddlepaddle.org.cn/packages/stable/cpu/
   pip install torch==2.7.0 torchvision==0.22.0
diff --git a/README_en.md b/README_en.md
index 1b532f4..888ee90 100755
--- a/README_en.md
+++ b/README_en.md
@@ -202,7 +202,7 @@ This project supports four running modes: CUDA (NVIDIA GPU acceleration), CPU (n
 ##### (4) Running on macOS (Apple Silicon)
 - Suitable for macOS (Apple Silicon) devices
 - For macOS (Intel), please use the CPU mode. Forcing GPU usage will only be slower.
-- On macOS (Apple Silicon), the accuracy of the subtitle detection precision mode model (ch_det_fast/model.onnx) does not seem to be ideal. It is recommended to use the fast mode instead.
+- On macOS (Apple Silicon), the accuracy of the PP-OCRv4-Server model for subtitle detection seems suboptimal. We recommend using an alternative model.
   ```shell
   pip install paddlepaddle==3.0.0 -i https://www.paddlepaddle.org.cn/packages/stable/cpu/
   pip install torch==2.7.0 torchvision==0.22.0
diff --git a/backend/config.py b/backend/config.py
index 2d11076..97357f5 100644
--- a/backend/config.py
+++ b/backend/config.py
@@ -52,7 +52,7 @@ class Config(QConfig):
     # 【设置inpaint算法】
     inpaintMode = OptionsConfigItem("Main", "InpaintMode", InpaintMode.STTN_AUTO, OptionsValidator(InpaintMode), EnumSerializer(InpaintMode))
     
-    subtitleDetectMode =  OptionsConfigItem("Main", "SubtitleDetectMode", SubtitleDetectMode.Accurate, OptionsValidator(SubtitleDetectMode), EnumSerializer(SubtitleDetectMode))
+    subtitleDetectMode =  OptionsConfigItem("Main", "SubtitleDetectMode", SubtitleDetectMode.PP_OCRv4_SERVER, OptionsValidator(SubtitleDetectMode), EnumSerializer(SubtitleDetectMode))
 
     # 【设置像素点偏差】
     # 用于判断是不是非字幕区域(一般认为字幕文本框的长度是要大于宽度的，如果字幕框的高大于宽，且大于的幅度超过指定像素点大小，则认为是错误检测)
diff --git a/backend/interface/ch.ini b/backend/interface/ch.ini
index 342a09c..228cbc6 100644
--- a/backend/interface/ch.ini
+++ b/backend/interface/ch.ini
@@ -76,7 +76,6 @@ DeleteABSection = 删除当前处理区块
 DeleteSelection = 删除当前激活选区
 
 [Main]
-SubtitleDetectionAcceleratorON = 字幕检测使用{}进行加速
 OnnxExectionProviderNotSupportedSkipped = ONNX 执行提供程序: {} 不支持，已跳过。
 OnnxExecutionProviderDetected=检测到 ONNX 执行提供程序: {}
 OnnxRuntimeNotInstall = ONNX 运行环境未安装，已跳过。
@@ -85,7 +84,8 @@ DirectMLWarning = 警告: DirectML 加速仅适用于 STTN 模型，其他模型
 ProcessingStartFindingSubtitles = [处理中] 开始查找字幕...
 FinishedFindingSubtitles = [结束] 查找字幕完成...
 ProcessingStartRemovingSubtitles = [处理中] 开始移除字幕...
-UseModel = 去除字幕使用模型: {}
+SubtitleRemoverModel = 去除字幕模型: {}
+SubtitleDetectionModel = 字幕检测模型: {}
 FullScreenProcessingNote = 未设置字幕区域，将对全屏进行处理，最终效果可能不理想
 ReadFileFailed = 读取文件 {} 失败
 FinishedProcessing = [完成] 字幕移除成功, 文件已保存到 {}
@@ -124,10 +124,6 @@ LAMA = LAMA
 ProPainter = ProPainter
 OpenCV = OpenCV
 
-[SubtitleDetectMode]
-Fast = 快速
-Accurate = 精准
-
 [InterfaceLanguage]
 ChineseSimplified = 简体中文
 ChineseTraditional = 繁體中文
diff --git a/backend/interface/chinese_cht.ini b/backend/interface/chinese_cht.ini
index ff42f9a..5a17064 100644
--- a/backend/interface/chinese_cht.ini
+++ b/backend/interface/chinese_cht.ini
@@ -76,7 +76,6 @@ DeleteABSection = 刪除當前處理區塊
 DeleteSelection = 刪除當前激活選區
 
 [Main]  
-SubtitleDetectionAcceleratorON = 字幕檢測使用{}進行加速
 OnnxExectionProviderNotSupportedSkipped = ONNX 執行提供者: {} 不支援，已略過。  
 OnnxExecutionProviderDetected = 偵測到 ONNX 執行提供者: {}  
 OnnxRuntimeNotInstall = ONNX 執行環境未安裝，已略過。  
@@ -85,7 +84,8 @@ DirectMLWarning = 警告：DirectML 加速僅適用於 STTN 模型，其他模
 ProcessingStartFindingSubtitles = [處理中] 開始搜尋字幕...  
 FinishedFindingSubtitles = [結束] 搜尋字幕完成...  
 ProcessingStartRemovingSubtitles = [處理中] 開始移除字幕...  
-UseModel = 去除字幕使用模型: {}
+SubtitleRemoverModel = 去除字幕模型: {}
+SubtitleDetectionModel = 字幕檢測模型: {}
 FullScreenProcessingNote = 未設定字幕區域，將對全螢幕進行處理，最終效果可能不理想  
 ReadFileFailed = 讀取檔案 {} 失敗  
 FinishedProcessing = [完成] 字幕移除成功，檔案已儲存至 {}  
@@ -124,10 +124,6 @@ LAMA = LAMA
 ProPainter = ProPainter  
 OpenCV = OpenCV  
 
-[SubtitleDetectMode]  
-Fast = 快速  
-Accurate = 精準  
-
 [InterfaceLanguage]  
 ChineseSimplified = 简体中文  
 ChineseTraditional = 繁體中文  
diff --git a/backend/interface/en.ini b/backend/interface/en.ini
index f440937..29886e9 100644
--- a/backend/interface/en.ini
+++ b/backend/interface/en.ini
@@ -76,7 +76,6 @@ DeleteABSection = Delete Section
 DeleteSelection = Delete Selection
 
 [Main]  
-SubtitleDetectionAcceleratorON = Subtitle detection accelerated with {}
 OnnxExectionProviderNotSupportedSkipped = ONNX provider: {} not supported, skipped.  
 OnnxExecutionProviderDetected = Detected ONNX provider: {}  
 OnnxRuntimeNotInstall = ONNX runtime not installed, skipped.  
@@ -85,7 +84,8 @@ DirectMLWarning = Warning: DirectML acceleration only works with STTN model.
 ProcessingStartFindingSubtitles = [Processing] Detecting subtitles...  
 FinishedFindingSubtitles = [Complete] Subtitle detection finished.  
 ProcessingStartRemovingSubtitles = [Processing] Removing subtitles...  
-UseModel = Use model for subtitle removal: {}
+SubtitleRemoverModel = Subtitle removal model: {}
+SubtitleDetectionModel = Subtitle detection model: {}
 FullScreenProcessingNote = Processing full screen (no area selected). Quality may vary.  
 ReadFileFailed = Failed to read file: {}  
 FinishedProcessing = [Complete] Subtitles removed. Output saved to: {}  
@@ -124,10 +124,6 @@ LAMA = LAMA
 ProPainter = ProPainter  
 OpenCV = OpenCV  
 
-[SubtitleDetectMode]  
-Fast = Fast  
-Accurate = Accurate  
-
 [InterfaceLanguage]  
 ChineseSimplified = 简体中文  
 ChineseTraditional = 繁體中文  
diff --git a/backend/interface/es.ini b/backend/interface/es.ini
index 0580c94..87b25aa 100644
--- a/backend/interface/es.ini
+++ b/backend/interface/es.ini
@@ -76,7 +76,6 @@ DeleteABSection = Eliminar sección
 DeleteSelection = Eliminar selección
 
 [Main]  
-SubtitleDetectionAcceleratorON = Detección de subtítulos acelerada con {}
 OnnxExectionProviderNotSupportedSkipped = Proveedor ONNX: {} no soportado, omitido.  
 OnnxExecutionProviderDetected = Proveedor ONNX detectado: {}  
 OnnxRuntimeNotInstall = Entorno ONNX no instalado, omitido.  
@@ -85,7 +84,8 @@ DirectMLWarning = Advertencia: Aceleración DirectML solo funciona con modelo ST
 ProcessingStartFindingSubtitles = [Procesando] Detectando subtítulos...  
 FinishedFindingSubtitles = [Completo] Detección finalizada.  
 ProcessingStartRemovingSubtitles = [Procesando] Eliminando subtítulos...  
-UseModel = Usar modelo para eliminar subtítulos: {}
+SubtitleRemoverModel = Modelo de eliminación de subtítulos: {}
+SubtitleDetectionModel = Modelo de detección de subtítulos: {}
 FullScreenProcessingNote = Procesando pantalla completa (área no seleccionada).  
 ReadFileFailed = Error al leer archivo: {}  
 FinishedProcessing = [Completo] Subtítulos eliminados. Guardado en: {}  
@@ -124,10 +124,6 @@ LAMA = LAMA
 ProPainter = ProPainter  
 OpenCV = OpenCV  
 
-[SubtitleDetectMode]  
-Fast = Rápido  
-Accurate = Preciso  
-
 [InterfaceLanguage]  
 ChineseSimplified = 简体中文  
 ChineseTraditional = 繁體中文  
diff --git a/backend/interface/japan.ini b/backend/interface/japan.ini
index b7ce6e9..58cc1ae 100644
--- a/backend/interface/japan.ini
+++ b/backend/interface/japan.ini
@@ -76,7 +76,6 @@ DeleteABSection = 現在の処理区間を削除
 DeleteSelection = 現在のアクティブ選択範囲を削除
 
 [Main]
-SubtitleDetectionAcceleratorON = 字幕検出を{}で加速
 OnnxExectionProviderNotSupportedSkipped = ONNXプロバイダ: {} 非対応
 OnnxExecutionProviderDetected=ONNXプロバイダ検出: {}
 OnnxRuntimeNotInstall = ONNXランタイム未インストール
@@ -85,7 +84,8 @@ DirectMLWarning = 警告: DirectML加速はSTTNモデルのみ
 ProcessingStartFindingSubtitles = [処理中] 字幕検索開始...
 FinishedFindingSubtitles = [完了] 字幕検索終了
 ProcessingStartRemovingSubtitles = [処理中] 字幕削除開始...
-UseModel = 字幕除去用モデル: {}
+SubtitleRemoverModel = 字幕除去モデル: {}
+SubtitleDetectionModel = 字幕検出モデル: {}
 FullScreenProcessingNote = 全画面処理（領域未選択）
 ReadFileFailed = ファイル読み込み失敗: {}
 FinishedProcessing = [完了] 字幕削除成功。保存先: {}
@@ -124,10 +124,6 @@ LAMA = LAMA
 ProPainter = ProPainter
 OpenCV = OpenCV
 
-[SubtitleDetectMode]
-Fast = 高速
-Accurate = 高精度
-
 [InterfaceLanguage]
 ChineseSimplified = 简体中文
 ChineseTraditional = 繁體中文
diff --git a/backend/interface/ko.ini b/backend/interface/ko.ini
index eac91d3..40b0eb8 100644
--- a/backend/interface/ko.ini
+++ b/backend/interface/ko.ini
@@ -76,7 +76,6 @@ DeleteABSection = 현재 처리 구간 삭제
 DeleteSelection = 현재 활성 선택 영역 삭제
 
 [Main]
-SubtitleDetectionAcceleratorON = 자막 감지 {} 가속 사용
 OnnxExectionProviderNotSupportedSkipped = ONNX 공급자: {} 지원 안됨
 OnnxExecutionProviderDetected=ONNX 공급자 감지: {}
 OnnxRuntimeNotInstall = ONNX 런타임 미설치
@@ -85,7 +84,8 @@ DirectMLWarning = 경고: DirectML 가속은 STTN 모델 전용
 ProcessingStartFindingSubtitles = [진행중] 자막 검색 시작...
 FinishedFindingSubtitles = [완료] 자막 검색 완료
 ProcessingStartRemovingSubtitles = [진행중] 자막 제거 시작...
-UseModel = 자막 제거 모델 사용: {}
+SubtitleRemoverModel = 자막 제거 모델: {}
+SubtitleDetectionModel = 자막 감지 모델: {}
 FullScreenProcessingNote = 전체 화면 처리 (영역 미선택)
 ReadFileFailed = 파일 읽기 실패: {}
 FinishedProcessing = [완료] 자막 제거 완료. 저장 위치: {}
@@ -124,10 +124,6 @@ LAMA = LAMA
 ProPainter = ProPainter
 OpenCV = OpenCV
 
-[SubtitleDetectMode]
-Fast = 빠름
-Accurate = 정확
-
 [InterfaceLanguage]
 ChineseSimplified = 简体中文
 ChineseTraditional = 繁體中文
diff --git a/backend/interface/vi.ini b/backend/interface/vi.ini
index 7eb8f17..6b86c63 100644
--- a/backend/interface/vi.ini
+++ b/backend/interface/vi.ini
@@ -76,7 +76,6 @@ DeleteABSection = Xóa vùng xử lý hiện tại
 DeleteSelection = Xóa vùng chọn hiện tại
 
 [Main]
-SubtitleDetectionAcceleratorON = Phát hiện phụ đề được tăng tốc bằng {}
 OnnxExectionProviderNotSupportedSkipped = ONNX provider: {} không hỗ trợ
 OnnxExecutionProviderDetected= Phát hiện ONNX provider: {}
 OnnxRuntimeNotInstall = Chưa cài đặt ONNX Runtime
@@ -85,7 +84,8 @@ DirectMLWarning = Cảnh báo: DirectML chỉ hỗ trợ STTN
 ProcessingStartFindingSubtitles = [Đang xử lý] Bắt đầu tìm phụ đề...
 FinishedFindingSubtitles = [Hoàn thành] Tìm phụ đề xong
 ProcessingStartRemovingSubtitles = [Đang xử lý] Bắt đầu xóa phụ đề...
-UseModel = Sử dụng mô hình xóa phụ đề: {}
+SubtitleRemoverModel = Mô hình xóa phụ đề: {}
+SubtitleDetectionModel = Mô hình phát hiện phụ đề: {}
 FullScreenProcessingNote = Xử lý toàn màn hình (không chọn vùng)
 ReadFileFailed = Lỗi đọc file: {}
 FinishedProcessing = [Hoàn thành] Xóa phụ đề thành công, lưu tại: {}
@@ -124,10 +124,6 @@ LAMA = LAMA
 ProPainter = ProPainter
 OpenCV = OpenCV
 
-[SubtitleDetectMode]
-Fast = Nhanh
-Accurate = Chính xác
-
 [InterfaceLanguage]
 ChineseSimplified = 简体中文
 ChineseTraditional = 繁體中文
diff --git a/backend/main.py b/backend/main.py
index e729f55..afe37b4 100644
--- a/backend/main.py
+++ b/backend/main.py
@@ -334,7 +334,6 @@ class SubtitleRemover:
         # 如果使用GPU加速，则打印GPU加速提示
         if self.hardware_accelerator.has_accelerator():
             accelerator_name = self.hardware_accelerator.accelerator_name
-            self.append_output(tr['Main']['SubtitleDetectionAcceleratorON'].format(accelerator_name))
             if accelerator_name == 'DirectML' and config.inpaintMode.value not in [InpaintMode.STTN_AUTO, InpaintMode.STTN_DET]:
                 self.append_output(tr['Main']['DirectMLWarning'])
         os.makedirs(os.path.dirname(self.video_out_path), exist_ok=True)
@@ -401,7 +400,8 @@ class SubtitleRemover:
                 model_device = 'DirectML'
             if self.hardware_accelerator.has_cuda() or self.hardware_accelerator.has_mps():
                 model_device = accelerator_name
-        self.append_output(tr['Main']['UseModel'].format(f"{model_friendly_name} ({model_device})"))
+        self.append_output(tr['Main']['SubtitleRemoverModel'].format(f"{model_friendly_name} ({model_device})"))
+        self.append_output(tr['Main']['SubtitleDetectionModel'].format(f"{config.subtitleDetectMode.value.value} ({", ".join(self.hardware_accelerator.onnx_providers)})"))
 
     def merge_audio_to_video(self):
         # 创建音频临时对象，windows下delete=True会有permission denied的报错
diff --git a/backend/models/V4/ch_det/fs_manifest.csv b/backend/models/V4/ch_det/fs_manifest.csv
new file mode 100644
index 0000000..8c89ff4
--- /dev/null
+++ b/backend/models/V4/ch_det/fs_manifest.csv
@@ -0,0 +1,4 @@
+filename,filesize,encoding,header

+inference_1.onnx,50000000,,

+inference_2.onnx,50000000,,

+inference_3.onnx,13443410,,

diff --git a/backend/models/V4/ch_det/inference_1.onnx b/backend/models/V4/ch_det/inference_1.onnx
new file mode 100644
index 0000000..3c85877
Binary files /dev/null and b/backend/models/V4/ch_det/inference_1.onnx differ
diff --git a/backend/models/V4/ch_det/inference_2.onnx b/backend/models/V4/ch_det/inference_2.onnx
new file mode 100644
index 0000000..1d70fb8
Binary files /dev/null and b/backend/models/V4/ch_det/inference_2.onnx differ
diff --git a/backend/models/V4/ch_det/inference_3.onnx b/backend/models/V4/ch_det/inference_3.onnx
new file mode 100644
index 0000000..9b394fa
Binary files /dev/null and b/backend/models/V4/ch_det/inference_3.onnx differ
diff --git a/backend/models/V4/ch_det_fast/inference.onnx b/backend/models/V4/ch_det_fast/inference.onnx
new file mode 100644
index 0000000..a4efe53
Binary files /dev/null and b/backend/models/V4/ch_det_fast/inference.onnx differ
diff --git a/backend/tools/constant.py b/backend/tools/constant.py
index 0645bc5..3eec677 100644
--- a/backend/tools/constant.py
+++ b/backend/tools/constant.py
@@ -16,5 +16,7 @@ class SubtitleDetectMode(Enum):
     """
     字幕检测算法枚举
     """
-    Fast = 0
-    Accurate = 1
\ No newline at end of file
+    PP_OCRv4_SERVER = "PP-OCRv4-Server" 
+    PP_OCRv4_MOBILE = "PP-OCRv4-Mobile"
+    PP_OCRv5_SERVER = "PP-OCRv5-Server" 
+    PP_OCRv5_MOBILE = "PP-OCRv5-Mobile" 
\ No newline at end of file
diff --git a/backend/tools/model_config.py b/backend/tools/model_config.py
index 5ee43c9..9c0107d 100644
--- a/backend/tools/model_config.py
+++ b/backend/tools/model_config.py
@@ -9,10 +9,14 @@ class ModelConfig:
         self.STTN_AUTO_MODEL_PATH = os.path.join(BASE_DIR, 'models', 'sttn-auto', 'infer_model.pth')
         self.STTN_DET_MODEL_PATH = os.path.join(BASE_DIR, 'models', 'sttn-det', 'sttn.pth')
         self.PROPAINTER_MODEL_DIR = os.path.join(BASE_DIR,'models', 'propainter')
-        if config.subtitleDetectMode.value == SubtitleDetectMode.Fast:
+        if config.subtitleDetectMode.value == SubtitleDetectMode.PP_OCRv5_MOBILE:
             self.DET_MODEL_DIR = os.path.join(BASE_DIR,'models', 'V5', 'ch_det_fast')
-        elif config.subtitleDetectMode.value == SubtitleDetectMode.Accurate:
+        elif config.subtitleDetectMode.value == SubtitleDetectMode.PP_OCRv5_SERVER:
             self.DET_MODEL_DIR = os.path.join(BASE_DIR, 'models', 'V5', 'ch_det')
+        elif config.subtitleDetectMode.value == SubtitleDetectMode.PP_OCRv4_MOBILE:
+            self.DET_MODEL_DIR = os.path.join(BASE_DIR,'models', 'V4', 'ch_det_fast')
+        elif config.subtitleDetectMode.value == SubtitleDetectMode.PP_OCRv4_SERVER:
+            self.DET_MODEL_DIR = os.path.join(BASE_DIR, 'models', 'V4', 'ch_det')
         else:
             raise ValueError(f"Invalid subtitle detect mode: {config.subtitleDetectMode.value}")
 
diff --git a/ui/setting_interface.py b/ui/setting_interface.py
index 67527da..d83bc28 100644
--- a/ui/setting_interface.py
+++ b/ui/setting_interface.py
@@ -43,7 +43,7 @@ class SettingInterface(QtWidgets.QVBoxLayout):
             title=tr["SubtitleExtractorGUI"]["SubtitleDetectMode"],
             content="",
             parent=parent,
-            texts=[list(tr['SubtitleDetectMode'].values())[i] for i,_ in enumerate(config.subtitleDetectMode.validator.options)],
+            texts=[mode.value for mode in config.subtitleDetectMode.validator.options],
         )
         self.addWidget(self.subtitle_detect_model_combo)