Update app.py

app.py

@@ -67,6 +67,26 @@ def get_indices_string(file_list, prefix_pattern):
             indices.append(match.group(1))
     return "".join(indices)
 
+def split_files_by_extension(all_files):
+    """
+    Split a list of files into videos (.mp4) and audios (.mp3, .wav) based on extensions.
+    Args:
+        all_files: List of file paths (e.g., ['file1.mp4', 'audio1.mp3', 'file2.mp4', 'audio2.mp3'])
+    Returns:
+        Tuple of (video_files, audio_files)
+    """
+    video_files = []
+    audio_files = []
+    for file_path in all_files:
+        filename = os.path.basename(file_path).lower()
+        if filename.endswith('.mp4'):
+            video_files.append(file_path)
+        elif filename.endswith(('.mp3', '.wav')):
+            audio_files.append(file_path)
+        else:
+            logger.warning(f"Unsupported file extension for {file_path}; skipping")
+    return video_files, audio_files
+
 def trim_silence(audio_clip, threshold=0.005):
     """
     Trim silence from the start and end of an audio clip.
@@ -130,11 +150,11 @@ def merge_videos_and_audios(video_files=None, audio_files=None, orig_vol=1.0, mu
     try:
         # Sort files by numerical index
         video_files = sort_files_by_index(video_files, r'file(\d+)\.mp4')
-        audio_files = sort_files_by_index(audio_files, r'audio(\d+)\.mp4')
+        audio_files = sort_files_by_index(audio_files, r'audio(\d+)\.(mp3|wav)')
         
         # Get indices for output naming
         video_indices = get_indices_string(video_files, r'file(\d+)\.mp4')
-        audio_indices = get_indices_string(audio_files, r'audio(\d+)\.mp4')
+        audio_indices = get_indices_string(audio_files, r'audio(\d+)\.(mp3|wav)')
 
         # Ensure at least two files are provided (videos, audios, or combination)
         video_count = len(video_files) if video_files else 0
@@ -302,21 +322,28 @@ def merge_videos_and_audios(video_files=None, audio_files=None, orig_vol=1.0, mu
 
 # --- Gradio App Using Blocks ---
 
-def gradio_merge_videos(video_files, audio_files, orig_vol, music_vol):
+def gradio_merge_files(all_files, orig_vol, music_vol):
     """
     Gradio endpoint to merge videos and/or audio.
     Args:
-        video_files: List of video file paths (optional)
-        audio_files: List of audio file paths (optional)
+        all_files: List of file paths (mixed videos and audios, e.g., ['file1.mp4', 'audio1.mp3', 'file2.mp4'])
         orig_vol: Volume for original video audio (0.0 to 1.0)
         music_vol: Volume for background audio (0.0 to 1.0)
     Returns:
         Path to the merged file (video or audio) or error message
     """
-    logger.info(f"Received {len(video_files) if video_files else 0} video files and {len(audio_files) if audio_files else 0} audio files")
+    logger.info(f"Received {len(all_files) if all_files else 0} files: {all_files}")
+    
+    # Split files into videos and audios based on extensions
+    video_files, audio_files = split_files_by_extension(all_files)
+    logger.info(f"Identified {len(video_files)} video files: {video_files}")
+    logger.info(f"Identified {len(audio_files)} audio files: {audio_files}")
     
     result = merge_videos_and_audios(
-        video_files, audio_files, orig_vol=orig_vol, music_vol=music_vol
+        video_files=video_files,
+        audio_files=audio_files,
+        orig_vol=orig_vol,
+        music_vol=music_vol
     )
     
     if isinstance(result, str) and result.startswith("Error"):
@@ -358,12 +385,11 @@ if __name__ == "__main__":
 
     with gr.Blocks(title="Video and Audio Merger API") as app:
         gr.Markdown("## Video and Audio Merger API")
-        gr.Markdown("Upload at least 2 files total (videos, audios, or a combination) to merge them. Videos and audios are optional.")
-        gr.Markdown("For API usage, name videos as file1.mp4, file2.mp4, etc., and audios as audio1.mp4, audio2.mp4, etc.")
+        gr.Markdown("Upload at least 2 files total (videos, audios, or a combination) to merge them.")
+        gr.Markdown("For API usage, name videos as file1.mp4, file2.mp4, etc., and audios as audio1.mp3, audio2.mp3, etc. Files can be sent in a single list; the app will distinguish them by extension (.mp4 for videos, .mp3/.wav for audios).")
         
         with gr.Row():
-            video_input = gr.File(label="Video Files (Optional, at least 2 if no audio)", type="filepath", file_count="multiple")
-            audio_input = gr.File(label="Audio Files (Optional, at least 2 if no video)", type="filepath", file_count="multiple")
+            file_input = gr.File(label="Upload Files (Videos: .mp4, Audios: .mp3/.wav)", type="filepath", file_count="multiple")
         
         with gr.Row():
             orig_vol_input = gr.Slider(minimum=0.0, maximum=1.0, value=1.0, step=0.05, label="Original Video Audio Volume")
@@ -374,8 +400,8 @@ if __name__ == "__main__":
         merge_button = gr.Button("Merge Files")
         
         merge_button.click(
-            fn=gradio_merge_videos,
-            inputs=[video_input, audio_input, orig_vol_input, music_vol_input],
+            fn=gradio_merge_files,
+            inputs=[file_input, orig_vol_input, music_vol_input],
             outputs=[output_video, output_audio]
         )