Update app_parallel.py

app_parallel.py

@@ -176,11 +176,11 @@ def custom_cleanup(temp_dir):
     import gc
     gc.collect()
 
-def get_audio_duration(audio_path):
-    audio_clip = mp.AudioFileClip(audio_path)
-    duration_in_seconds = audio_clip.duration
-    audio_clip.close()  # Don't forget to close the clip
-    return duration_in_seconds
+# def get_audio_duration(audio_path):
+#     audio_clip = mp.AudioFileClip(audio_path)
+#     duration_in_seconds = audio_clip.duration
+#     audio_clip.close()  # Don't forget to close the clip
+#     return duration_in_seconds
 
 
 def generate_audio(voice_cloning, voice_gender, text_prompt):
@@ -227,10 +227,10 @@ def generate_audio(voice_cloning, voice_gender, text_prompt):
                 temp_file.write(chunk)
             driven_audio_path = temp_file.name
             print('driven_audio_path',driven_audio_path)
-            audio_duration = get_audio_duration(driven_audio_path)
-            print('Total Audio Duration in seconds',audio_duration)
+            # audio_duration = get_audio_duration(driven_audio_path)
+            # print('Total Audio Duration in seconds',audio_duration)
 
-    return driven_audio_path, audio_duration
+    return driven_audio_path
 
 def run_preprocessing(args):
     global path_of_lm_croper, path_of_net_recon_model, dir_of_BFM_fitting
@@ -262,20 +262,26 @@ def openai_chat_avatar(text_prompt):
     )
     return response
     
-def split_audio(audio_path, chunk_duration):
+def split_audio(audio_path, TEMP_DIR, chunk_duration):
     audio_clip = mp.AudioFileClip(audio_path)
     total_duration = audio_clip.duration
-
+    print("split_audio duration:",total_duration)
+    number_of_chunks = math.ceil(total_duration / chunk_duration)
+    print("Number of audio chunks:",number_of_chunks)
     audio_chunks = []
-    for start_time in range(0, int(total_duration), chunk_duration):
+    for i in range(number_of_chunks):
+        start_time = i * chunk_duration
         end_time = min(start_time + chunk_duration, total_duration)
         chunk = audio_clip.subclip(start_time, end_time)
+        
+        # Create a temporary file for the chunk
         with tempfile.NamedTemporaryFile(suffix=f"_chunk_{start_time}-{end_time}.wav", prefix="audio_chunk_", dir=TEMP_DIR.name, delete=False) as temp_file:
             chunk_path = temp_file.name
-            chunk.write_audiofile(chunk_path)
+            chunk.write_audiofile(chunk_path)  # Specify codec if needed
             audio_chunks.append((start_time, chunk_path))
     
-    return audio_chunks
+    audio_clip.close()  # Close the audio clip to release resources
+    return audio_chunks, total_duration
 
 # Generator function to yield chunk results as they are processed
 def generate_chunks(audio_chunks, preprocessed_data, args):
@@ -344,7 +350,7 @@ def parallel_processing():
             source_image_path = save_uploaded_file(source_image, 'source_image.png',TEMP_DIR)
             print(source_image_path)
     
-            driven_audio_path, audio_duration = generate_audio(voice_cloning, voice_gender, text_prompt)
+            driven_audio_path = generate_audio(voice_cloning, voice_gender, text_prompt)
         
             save_dir = tempfile.mkdtemp(dir=TEMP_DIR.name)
             result_folder = os.path.join(save_dir, "results")
@@ -355,7 +361,7 @@ def parallel_processing():
             preprocessed_data = run_preprocessing(args)
             chunk_duration = 3
             print(f"Splitting the audio into {chunk_duration}-second chunks...")
-            audio_chunks = split_audio(driven_audio_path, chunk_duration=chunk_duration)
+            audio_chunks, audio_duration = split_audio(driven_audio_path, TEMP_DIR, chunk_duration=chunk_duration)
             print(f"Audio has been split into {len(audio_chunks)} chunks: {audio_chunks}")
 
             os.makedirs('lives', exist_ok=True)