edited vc to support webm input

Dockerfile

@@ -25,6 +25,9 @@ RUN chmod -R 777 /app/results_st
 RUN mkdir /app/results_vtr
 RUN chmod -R 777 /app/results_vtr
 
+RUN mkdir /app/logs
+RUN chmod -R 777 /app/logs
+
 COPY . .
 
 CMD ["python", "app.py"]

app.py

@@ -38,7 +38,7 @@ def classify_input_format(input_string):
         pass
     
     # if the input format is neither seconds nor HH:MM:SS, return None
-    gradio.Error("Input format not supported. Please enter a valid time in the format HH:MM:SS or seconds.")
+    gr.Error("Input format not supported. Please enter a valid time in the format HH:MM:SS or seconds.")
 
 def set_output_file(input_file, output_format, folder, insert_string = 'trimmed'):
     # Set output file name and extension
@@ -183,7 +183,8 @@ def convert_video(input_file, output_format):
     try:
         # Set output file
         output_folder = f"{os.getcwd()}/results_vc/"
-        output_file = set_output_file(input_file, output_format, output_folder)
+        output_file = set_output_file(input_file, output_format, output_folder, \
+                                      insert_string = 'converted')
 
         # Convert video
         output_file = convert_video_helper(input_file, output_file, output_format)

utlis.py

@@ -7,6 +7,11 @@ from pathlib import Path
 import gradio as gr
 import pandas as pd
 
+import logging
+os.makedirs(f'{os.getcwd()}/logs', exist_ok=True)
+logging.basicConfig(filename=f'{os.getcwd()}/logs/logfile.log', level=logging.INFO, 
+                    format='%(asctime)s - %(levelname)s - %(message)s')
+logging.info('Starting the application...')
 
 def sort_transcript(file_path, save_path):
     """
@@ -55,41 +60,88 @@ def sort_transcript(file_path, save_path):
 
 
 def convert_video_format(media_in, media_out):
+    """
+      Function to convert video format using ffmpeg.
+      
+    """
+    logging.info(f'...Converting video format from {media_in} to {media_out}...')
+
     try:
         WAV_CHANNELS = 1
         WAV_SAMPLE_RATE = 16000
-        
+
         if not isinstance(media_in, (str, Path)):
             raise TypeError("media_in must be a string or a PathLike object")
-        
+
         if not isinstance(media_out, (str, Path)):
             raise TypeError("media_out must be a string or a PathLike object")
-        ext = Path(media_out).suffix
-        if ext == '.wav':
-            # convert to wav with standard format for audio models         
-            print(f'...generating {media_out}...')            
-            subprocess.run(f'ffmpeg -y -i "{media_in}" -acodec pcm_s16le -ac {WAV_CHANNELS} -ar {WAV_SAMPLE_RATE} "{media_out}" -hide_banner -loglevel warning', check=True, shell=True)
-            return media_out
 
-        else:
-            print(f'...Using ffmpeg to convert {media_in} to {media_out}...')      
-            print(f'...generating {media_out}...')            
+        in_ext = Path(media_in).suffix
+        out_ext = Path(media_out).suffix
+
+        if in_ext !='.webm':
+          # input is not in webm format
+          if out_ext == '.wav':
+              # convert to wav with standard format for audio models
+              logging.info(f'...generating {media_out}...')
+              command = ['ffmpeg', '-y', '-i', 
+                          media_in, '-acodec', 'pcm_s16le', 
+                          '-ac', str(WAV_CHANNELS), '-ar', 
+                          str(WAV_SAMPLE_RATE), media_out, 
+                          '-hide_banner', '-loglevel', 'warning']
+              process = subprocess.run(command, capture_output=True, text=True)
+              if process.returncode != 0:
+                logging.info(f"Error: {process.stderr}")
+              else:
+                logging.info(process.stdout)
+              return media_out
+
+          else:
+              logging.info(f'...Using ffmpeg to convert {media_in} to {media_out}...')
+              logging.info(f'...generating {media_out}...')
+              command = ['ffmpeg',
+                              '-y',
+                              '-i',
+                              media_in,
+                              '-c',
+                              'copy',
+                              media_out,
+                              '-hide_banner',
+                              '-loglevel',
+                              'warning'
+                              ]
+              process = subprocess.run(command, check=True)
+              if process.returncode != 0:
+                logging.info(f"Error: {process.stderr}")
+              else:
+                logging.info(process.stdout)
+              return media_out
         
-            subprocess.run(['ffmpeg',
-                            '-y',
-                            '-i',
-                            media_in,
-                            '-c',
-                            'copy',
-                            media_out,
-                            '-hide_banner', 
-                            '-loglevel', 
-                            'warning'
-                            ], check=True)
+        if in_ext == '.webm':
+          if out_ext == '.wav':
+            command = ['ffmpeg', '-i', media_in, media_out]
+            process = subprocess.run(command, capture_output=True, text=True)
+            if process.returncode != 0:
+                logging.info(f"Error: {process.stderr}")
+            else:
+                logging.info(process.stdout)
+            return media_out
+          
+          else:
+            command = ['ffmpeg', '-i', media_in, '-vcodec', 'h264', '-acodec', 'aac', media_out]
+            process = subprocess.run(command, capture_output=True, text=True)
+            if process.returncode != 0:
+                logging.info(f"Error: {process.stderr}")
+            else:
+                logging.info(process.stdout)
             return media_out
+        
+            
     except Exception as e:
+        logging.error(f"Error converting video format: {e}")
         gr.Error(f"Error converting video format: {e}")
 
+
 def trim_media(media_in, media_out, start, end): 
     try: 
         # options for writing out audio if converting