Update app.py

app.py

@@ -8,6 +8,7 @@ import logging
 import sys
 import traceback
 import socket
+import shutil
 
 # Set up logging to debug issues
 logging.basicConfig(level=logging.INFO, handlers=[logging.StreamHandler(sys.stdout)])
@@ -132,7 +133,7 @@ def trim_silence(audio_clip, threshold=0.005):
         logger.error(f"Error trimming silence: {str(e)}")
         return audio_clip
 
-def merge_videos_and_audios(video_files=None, audio_files=None, orig_vol=1.0, music_vol=0.5):
+def merge_videos_and_audios(video_files=None, audio_files=None, orig_vol=1.0, music_vol=0.5, temp_dir=None):
     """
     Merge multiple video clips and/or audio clips based on inputs provided.
     - If only video_files: Merge videos, retaining their original audio.
@@ -144,6 +145,7 @@ def merge_videos_and_audios(video_files=None, audio_files=None, orig_vol=1.0, mu
         audio_files: List of audio file paths (optional)
         orig_vol: Volume for original video audio (0.0 to 1.0)
         music_vol: Volume for background audio (0.0 to 1.0)
+        temp_dir: Temporary directory to clean up (optional)
     Returns:
         Path to the merged file (video or audio) or error message.
     """
@@ -168,13 +170,15 @@ def merge_videos_and_audios(video_files=None, audio_files=None, orig_vol=1.0, mu
             logger.error(error_msg)
             return error_msg
 
-        # Create a temporary output path
-        temp_dir = tempfile.mkdtemp()
+        # Create a temporary output path (use provided temp_dir if available)
+        if temp_dir is None:
+            temp_dir = tempfile.mkdtemp()
+        output_dir = temp_dir
 
         # Case 1: Audio only
         if audio_count >= 2 and video_count == 0:
             output_filename = f"combined_audio_{audio_indices}.mp3"
-            output_path = os.path.join(temp_dir, output_filename)
+            output_path = os.path.join(output_dir, output_filename)
             logger.info("Merging audio files only")
             
             # Load, normalize, and trim audio clips
@@ -228,7 +232,7 @@ def merge_videos_and_audios(video_files=None, audio_files=None, orig_vol=1.0, mu
             output_filename = f"combined_video_{video_indices}_with_audio_{audio_indices}.mp4"
         else:
             output_filename = f"combined_video_{video_indices}.mp4"
-        output_path = os.path.join(temp_dir, output_filename)
+        output_path = os.path.join(output_dir, output_filename)
         
         # Load and concatenate video clips
         video_clips = [VideoFileClip(video) for video in video_files]
@@ -319,43 +323,94 @@ def merge_videos_and_audios(video_files=None, audio_files=None, orig_vol=1.0, mu
         error_msg = f"Error during merging: {str(e)}\n{traceback.format_exc()}"
         logger.error(error_msg)
         return error_msg
+    finally:
+        # Clean up temporary directory if it was created in this function
+        if temp_dir and os.path.exists(temp_dir):
+            try:
+                shutil.rmtree(temp_dir)
+                logger.info(f"Cleaned up temporary directory: {temp_dir}")
+            except Exception as e:
+                logger.warning(f"Failed to clean up temporary directory {temp_dir}: {str(e)}")
 
 # --- Gradio App Using Blocks ---
 
-def gradio_merge_files(all_files, orig_vol, music_vol):
+def gradio_merge_files(file_objects, orig_vol, music_vol):
     """
-    Gradio endpoint to merge videos and/or audio.
+    Gradio endpoint to merge videos and/or audio from binary file uploads.
     Args:
-        all_files: List of file paths (mixed videos and audios, e.g., ['file1.mp4', 'audio1.mp3', 'file2.mp4'])
+        file_objects: List of file objects (binary uploads)
         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(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=video_files,
-        audio_files=audio_files,
-        orig_vol=orig_vol,
-        music_vol=music_vol
-    )
-    
-    if isinstance(result, str) and result.startswith("Error"):
-        logger.error(result)
-        return result, None
-    else:
-        logger.info(f"Merge successful. Output saved at: {result}")
-        # Return appropriate output based on file type
-        if result.endswith(".mp3"):
-            return None, result  # Audio output
+    try:
+        logger.info(f"Received {len(file_objects) if file_objects else 0} file objects")
+        
+        if not file_objects or len(file_objects) < 2:
+            error_msg = "Error: Please upload at least 2 files."
+            logger.error(error_msg)
+            return error_msg, None
+
+        # Create a temporary directory to store uploaded files
+        temp_dir = tempfile.mkdtemp()
+        all_files = []
+
+        # Save each file object to the temporary directory
+        for file_obj in file_objects:
+            if file_obj is None:
+                logger.warning("Skipping None file object")
+                continue
+            # Get the original filename from the file object
+            original_filename = os.path.basename(file_obj.name)
+            # Ensure the filename matches expected patterns
+            if not re.match(r'file\d+\.mp4', original_filename, re.IGNORECASE) and \
+               not re.match(r'audio\d+\.(mp3|wav)', original_filename, re.IGNORECASE):
+                logger.warning(f"Filename {original_filename} does not match expected pattern; skipping")
+                continue
+            # Create a temporary file path
+            temp_file_path = os.path.join(temp_dir, original_filename)
+            # Copy the file data to the temporary location
+            with open(temp_file_path, 'wb') as temp_file:
+                shutil.copyfileobj(file_obj, temp_file)
+            all_files.append(temp_file_path)
+            logger.info(f"Saved uploaded file to {temp_file_path}")
+
+        if len(all_files) < 2:
+            error_msg = "Error: Fewer than 2 valid files after filtering."
+            logger.error(error_msg)
+            return error_msg, None
+
+        # 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=video_files,
+            audio_files=audio_files,
+            orig_vol=orig_vol,
+            music_vol=music_vol,
+            temp_dir=temp_dir
+        )
+        
+        if isinstance(result, str) and result.startswith("Error"):
+            logger.error(result)
+            return result, None
         else:
-            return result, None  # Video output
+            logger.info(f"Merge successful. Output saved at: {result}")
+            # Return appropriate output based on file type
+            if result.endswith(".mp3"):
+                return None, result  # Audio output
+            else:
+                return result, None  # Video output
+    except Exception as e:
+        error_msg = f"Error processing files: {str(e)}\n{traceback.format_exc()}"
+        logger.error(error_msg)
+        return error_msg, None
+    finally:
+        # Cleanup is handled in merge_videos_and_audios
+        pass
 
 # --- Main Execution ---
 
@@ -386,10 +441,10 @@ 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.")
-        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).")
+        gr.Markdown("For API usage, send binary files via multipart/form-data. Name videos as file1.mp4, file2.mp4, etc., and audios as audio1.mp3, audio2.mp3, etc.")
         
         with gr.Row():
-            file_input = gr.File(label="Upload Files (Videos: .mp4, Audios: .mp3/.wav)", type="filepath", file_count="multiple")
+            file_input = gr.File(label="Upload Files (Videos: .mp4, Audios: .mp3/.wav)", type="file", 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")