Update app.py

app.py

@@ -133,33 +133,44 @@ def infer(audio_input_path):
 def load_chosen_audio(audio_path):
     return audio_path
 
-def overlay_audio(original_mp3: str, translated_wav: str, volume_reduction_db: int = 10) -> str:
+def overlay_audio(
+    original_mp3: str,
+    translated_wav: str,
+    volume_reduction_db: int = 10,
+    cut_start: float = 0.0
+) -> str:
     """
     Overlays translated audio on top of the original, reduces the original volume,
     and ensures the final audio lasts as long as the longer of the two tracks.
-
+    
     :param original_mp3: Path to the original MP3 file.
     :param translated_wav: Path to the translated WAV file.
     :param volume_reduction_db: Volume reduction in dB (default is -10 dB).
+    :param cut_start: Number of seconds to trim from the start of the translated audio (default: 0.0).
     :return: Path to the temporary output WAV file.
     """
     # Load original MP3 and convert to WAV
     original = AudioSegment.from_mp3(original_mp3).set_frame_rate(16000).set_channels(1)
     
-    # Lower the volume (default is -10 dB)
+    # Lower the volume
     original = original - volume_reduction_db
     
     # Load the translated WAV
     translated = AudioSegment.from_wav(translated_wav).set_frame_rate(16000).set_channels(1)
-    
+
+    # Trim the start of the translated audio if needed
+    if cut_start > 0:
+        cut_ms = int(cut_start * 1000)  # Convert seconds to milliseconds
+        translated = translated[cut_ms:]
+
     # Determine the final length (longer of the two)
     final_length = max(len(original), len(translated))
-    
+
     # Extend the shorter track with silence to match the longer track
     if len(original) < final_length:
-        original = original + AudioSegment.silent(duration=final_length - len(original))
+        original += AudioSegment.silent(duration=final_length - len(original))
     if len(translated) < final_length:
-        translated = translated + AudioSegment.silent(duration=final_length - len(translated))
+        translated += AudioSegment.silent(duration=final_length - len(translated))
 
     # Overlay the translated speech over the original
     combined = original.overlay(translated)
@@ -169,11 +180,10 @@ def overlay_audio(original_mp3: str, translated_wav: str, volume_reduction_db: i
     combined.export(temp_file.name, format="wav")
 
     print(f"Final audio saved at: {temp_file.name}")
-    return temp_file.name  # Return the temporary file path
 
-def process_final_combination(audio_in, chosen_translated, video_input):
+def process_final_combination(audio_in, chosen_translated, volume, cut_start, video_input):
     audio_in = process_audio(audio_in)
-    temp_output_path = overlay_audio(audio_in, chosen_translated)
+    temp_output_path = overlay_audio(audio_in, chosen_translated, volume, cut_start)
 
     if video_input:
         return gr.update(value=temp_output_path, visible=True), gr.update(visible=True)
@@ -251,6 +261,9 @@ with gr.Blocks(css=css) as demo:
                 scale=2
             )
             choose_this_btn = gr.Button("Use this one", scale=1, visible=False)
+        with gr.Row():
+            volume_reduction = gr.Slider(label="Volume reduction", minimum=0, maximum=12, step=1, value=12)
+            cut_start = gr.Slider(label="Cut start", minimum=0.0, maximum=4.0, step=0.1, value=2.0)
         combined_output = gr.Audio("Combined Output", type="filepath", visible=False)
         apply_to_video_btn = gr.Button("Apply to video", visible=False)
         final_video_out = gr.Video(label="Video + Translated Audio", visible=False)
@@ -281,7 +294,7 @@ with gr.Blocks(css=css) as demo:
 
     choose_this_btn.click(
         fn = process_final_combination,
-        inputs = [audio_input, dropdown_wav_selector, video_input],
+        inputs = [audio_input, dropdown_wav_selector, volume_reduction, cut_start, video_input],
         outputs = [combined_output, apply_to_video_btn]
     )