Update app.py

app.py

@@ -2,6 +2,7 @@ import gradio as gr
 import tempfile
 import os
 from moviepy.editor import VideoFileClip, concatenate_videoclips, AudioFileClip, CompositeAudioClip, concatenate_audioclips
+import numpy as np
 import logging
 import sys
 import traceback
@@ -25,6 +26,42 @@ def check_port(port):
         except socket.error:
             return False
 
+def trim_silence(audio_clip, threshold=0.01):
+    """
+    Trim silence from the start and end of an audio clip.
+    Args:
+        audio_clip: AudioFileClip object
+        threshold: Amplitude threshold below which audio is considered silent
+    Returns:
+        Trimmed AudioFileClip
+    """
+    # Get audio data as numpy array
+    samples = audio_clip.to_soundarray(fps=44100)
+    # Compute amplitude (RMS)
+    if len(samples.shape) > 1:  # Stereo audio
+        amplitudes = np.sqrt(np.mean(samples**2, axis=1))
+    else:  # Mono audio
+        amplitudes = np.sqrt(samples**2)
+    
+    # Find non-silent regions
+    non_silent = amplitudes > threshold
+    if not np.any(non_silent):
+        logger.warning("Audio clip is completely silent")
+        return audio_clip
+    
+    # Find start and end indices
+    start_idx = np.argmax(non_silent)
+    end_idx = len(non_silent) - np.argmax(non_silent[::-1])
+    
+    # Convert indices to time (seconds)
+    start_time = start_idx / 44100
+    end_time = end_idx / 44100
+    
+    # Trim the audio
+    trimmed_audio = audio_clip.subclip(start_time, end_time)
+    logger.info(f"Trimmed audio from {start_time:.2f}s to {end_time:.2f}s (original duration: {audio_clip.duration:.2f}s)")
+    return trimmed_audio
+
 def merge_videos_and_audios(video_files=None, audio_files=None, orig_vol=1.0, music_vol=0.5):
     """
     Merge multiple video clips and/or audio clips based on inputs provided.
@@ -60,9 +97,21 @@ def merge_videos_and_audios(video_files=None, audio_files=None, orig_vol=1.0, mu
             output_path = os.path.join(temp_dir, "merged_output.mp3")
             logger.info("Merging audio files only")
             
-            # Load and concatenate audio clips
-            audio_clips = [AudioFileClip(audio) for audio in audio_files]
+            # Load, normalize, and trim audio clips
+            audio_clips = []
+            for audio in audio_files:
+                clip = AudioFileClip(audio).set_fps(44100)  # Normalize sample rate
+                logger.info(f"Original audio duration for {audio}: {clip.duration:.2f}s")
+                trimmed_clip = trim_silence(clip)
+                audio_clips.append(trimmed_clip)
+            
+            # Log durations after trimming
+            for i, clip in enumerate(audio_clips):
+                logger.info(f"Trimmed audio {i+1} duration: {clip.duration:.2f}s")
+            
+            # Concatenate audio clips
             final_audio_clip = concatenate_audioclips(audio_clips)
+            logger.info(f"Concatenated audio duration: {final_audio_clip.duration:.2f}s")
             
             # Write the final audio
             logger.info(f"Writing output audio to {output_path}")
@@ -85,21 +134,35 @@ def merge_videos_and_audios(video_files=None, audio_files=None, orig_vol=1.0, mu
         
         # Determine final video duration
         video_duration = final_video_clip.duration or sum(clip.duration for clip in video_clips)
-        logger.info(f"Total video duration: {video_duration} seconds")
+        logger.info(f"Total video duration: {video_duration:.2f}s")
         
         # Handle audio (if provided)
         if audio_files:
             logger.info("Processing audio files")
-            # Load and concatenate audio clips
-            audio_clips = [AudioFileClip(audio) for audio in audio_files]
+            # Load, normalize, and trim audio clips
+            audio_clips = []
+            for audio in audio_files:
+                clip = AudioFileClip(audio).set_fps(44100)  # Normalize sample rate
+                logger.info(f"Original audio duration for {audio}: {clip.duration:.2f}s")
+                trimmed_clip = trim_silence(clip)
+                audio_clips.append(trimmed_clip)
+            
+            # Log durations after trimming
+            for i, clip in enumerate(audio_clips):
+                logger.info(f"Trimmed audio {i+1} duration: {clip.duration:.2f}s")
+            
+            # Concatenate audio clips
             concatenated_audio = concatenate_audioclips(audio_clips)
+            logger.info(f"Concatenated audio duration: {concatenated_audio.duration:.2f}s")
             
             # Adjust concatenated audio duration to match video duration (trim or loop)
             if concatenated_audio.duration > video_duration:
                 concatenated_audio = concatenated_audio.subclip(0, video_duration)
+                logger.info(f"Trimmed concatenated audio to match video duration: {concatenated_audio.duration:.2f}s")
             elif concatenated_audio.duration < video_duration:
                 # Loop the audio to match video duration
                 concatenated_audio = concatenated_audio.fx(lambda clip: clip.loop(duration=video_duration))
+                logger.info(f"Looped concatenated audio to match video duration: {concatenated_audio.duration:.2f}s")
             
             # Apply volume to concatenated audio
             concatenated_audio = concatenated_audio.volumex(music_vol)