添加文案处理的功能

resources/scripts/Lai.py

@@ -12,13 +12,13 @@ import shotSplit
 sys.stdout = io.TextIOWrapper(sys.stdout.buffer, encoding="utf-8")
 
 # 判断sys.argv 的长度，如果小于2，说明没有传入参数，设置初始参数
+# "C:\\Users\\27698\\Desktop\\LAITool\\resources\\scripts\\Lai.exe" -c "D:/来推项目集/7.4/娱乐：江湖大哥退休，去拍电影/scripts/output_crop_00001.json" "NVIDIA"
 if len(sys.argv) < 2:
     sys.argv = [
         "C:\\Users\\27698\\Desktop\\LAITool\\resources\\scripts\\Lai.exe",
-        "-ka",
-        "C:\\Users\\27698\\Desktop\\测试\\123\\测试用 不删.mp4",
-        "C:\\Users\\27698\\Desktop\\测试\\123\\测试用 不删.json",
-        30,
+        "-c",
+        "D:/来推项目集/7.4/娱乐：江湖大哥退休，去拍电影/scripts/output_crop_00001.json",
+        "NVIDIA",
     ]
 
 print(sys.argv)
@@ -91,3 +91,8 @@ elif sys.argv[1] == "-ka":
 elif sys.argv[1] == "-a":
     print("开始算法分镜：" + sys.argv[2] + " -- 输出文件夹：" + sys.argv[3])
     shotSplit.init(sys.argv[2], sys.argv[3], sys.argv[4], sys.argv[5], sys.argv[6])
+# 本地提取音频
+elif sys.argv[1] == "-t":
+    print("开始提取文字：" + sys.argv[2])
+    shotSplit.GetTextTask(sys.argv[2], sys.argv[3], sys.argv[4])
+    pass

resources/scripts/shotSplit.py

@@ -229,6 +229,58 @@ def GetText(out_folder, mp3_list):
     sys.stdout.flush()
 
 
+def GetTextTask(out_folder, mp, name):
+    text = []
+    # 先获取模型
+    print("正在下载或加载模型")
+    sys.stdout.flush()
+    model_path = Path(
+        hf_hub_download(repo_id="Systran/faster-whisper-large-v3", filename="model.bin")
+    )
+    hf_hub_download(
+        repo_id="Systran/faster-whisper-large-v3",
+        filename="config.json",
+    )
+    hf_hub_download(
+        repo_id="Systran/faster-whisper-large-v3",
+        filename="preprocessor_config.json",
+    )
+    hf_hub_download(
+        repo_id="Systran/faster-whisper-large-v3",
+        filename="tokenizer.json",
+    )
+    hf_hub_download(
+        repo_id="Systran/faster-whisper-large-v3",
+        filename="vocabulary.json",
+    )
+    model = WhisperModel(
+        model_size_or_path=os.path.dirname(model_path),
+        device="auto",
+        local_files_only=True,
+    )
+    print("模型加载成功，开始识别")
+    sys.stdout.flush()
+    segments, info = model.transcribe(
+        mp,
+        beam_size=5,
+        language="zh",
+        vad_filter=True,
+        vad_parameters=dict(min_silence_duration_ms=1000),
+    )
+    tmp_text = ""
+    for segment in segments:
+        tmp_text += segment.text + "。"
+    print(mp + "识别完成")
+    sys.stdout.flush()
+    text.append(tmp_text)
+
+    # 数据写出
+    sys.stdout.flush()
+    tools = public_tools.PublicTools()
+    tools.write_to_file(text, os.path.join(out_folder, name + ".txt"))
+    sys.stdout.flush()
+
+
 def get_fram(video_path, out_path, sensitivity):
     try:
         shijian_list = find_scenes(video_path, sensitivity)  # 多组时间列表