feat: add trim at the end of audio

app.py

@@ -3,7 +3,7 @@ import re
 import subprocess
 import os
 import shutil
-from pydub import AudioSegment
+from pydub import AudioSegment, silence
 import gradio as gr
 import traceback
 import logging
@@ -38,7 +38,7 @@ def handle_file_upload(file):
     shutil.copy(file.name, input_path)
     
     return input_path, formatted_title
-    
+
 def run_inference(model_type, config_path, start_check_point, input_dir, output_dir, device_ids="0"):
     command = [
         "python", "inference.py",
@@ -93,8 +93,6 @@ def move_stems_to_parent(input_dir):
                 new_instrumental_path = os.path.join(parent_dir, "instrumental.wav")
                 print(f"Moving {instrumental_path} to {new_instrumental_path}")
                 shutil.move(instrumental_path, new_instrumental_path)
-            else:
-                print(f"Instrumental file not found: {instrumental_path}")
 
 def combine_stems_for_all(input_dir):
     for subdir, _, _ in os.walk(input_dir):
@@ -118,10 +116,29 @@ def combine_stems_for_all(input_dir):
         stems = {name: AudioSegment.from_file(path) for name, path in stem_paths.items()}
         combined = stems["vocals"].overlay(stems["bass"]).overlay(stems["others"]).overlay(stems["instrumental"])
 
+        # Trim silence from the end of the combined audio
+        trimmed_combined = trim_silence_at_end(combined)
+
         output_file = os.path.join(subdir, f"{song_name}.MDS.wav")
-        combined.export(output_file, format="wav")
+        trimmed_combined.export(output_file, format="wav")
         print(f"Exported combined stems to {output_file}")
 
+def trim_silence_at_end(audio_segment, silence_thresh=-50, chunk_size=10):
+    """
+    Trims silence at the end of an AudioSegment.
+    :param audio_segment: The audio segment to trim.
+    :param silence_thresh: The threshold in dB below which is considered silence.
+    :param chunk_size: The size of the chunks in milliseconds that are checked for silence.
+    :return: A trimmed AudioSegment with silence removed from the end.
+    """
+    silence_end = silence.detect_silence(audio_segment, min_silence_len=chunk_size, silence_thresh=silence_thresh)
+    
+    if silence_end:
+        last_silence_start = silence_end[-1][0]
+        return audio_segment[:last_silence_start]
+    else:
+        return audio_segment
+
 def delete_folders_and_files(input_dir):
     folders_to_delete = ['htdemucs', 'mel_band_roformer', 'scnet', 'bs_roformer']
     files_to_delete = ['bass.wav', 'vocals.wav', 'other.wav', 'instrumental.wav']