Upload myinfer_latest.py

myinfer_latest.py

@@ -25,7 +25,7 @@ from pydub import AudioSegment
 import uuid
 from threading import Semaphore
 from threading import Lock
-
+from multiprocessing import Process, Queue
 
 
 
@@ -84,6 +84,8 @@ if os.path.isfile("rmvpe.pt"):
     f0method_mode.insert(2, "rmvpe")
     f0method_info = "PM is fast, Harvest is good but extremely slow, Rvmpe is alternative to harvest (might be better), and Crepe effect is good but requires GPU (Default: PM)"
 
+
+
 def load_hubert():
     global hubert_model
     models, _, _ = checkpoint_utils.load_model_ensemble_and_task(
@@ -168,7 +170,7 @@ def api_convert_voice():
         file = request.files['file']
         if file.filename == '':
             return jsonify({"error": "No selected file"}), 400
-        created_files = []
+        #created_files = []
         # Save the file to a temporary path
         unique_id = str(uuid.uuid4())
         print(unique_id)
@@ -177,7 +179,7 @@ def api_convert_voice():
         input_audio_path = os.path.join(tmp, f"{spk_id}_input_audio_{unique_id}.{filename.split('.')[-1]}")
         file.save(input_audio_path)
         
-        created_files.append(input_audio_path)
+        #created_files.append(input_audio_path)
         
         #split audio
         cut_vocal_and_inst(input_audio_path,spk_id,unique_id)
@@ -185,7 +187,16 @@ def api_convert_voice():
         vocal_path = f"output/{spk_id}_{unique_id}/{split_model}/{spk_id}_input_audio_{unique_id}/vocals.wav"
         inst = f"output/{spk_id}_{unique_id}/{split_model}/{spk_id}_input_audio_{unique_id}/no_vocals.wav"
         print("*****before making call to convert ", unique_id)
-        output_path = convert_voice(spk_id, vocal_path, voice_transform,unique_id)
+        output_queue = Queue()
+    
+    # Create and start the process
+        p = Process(target=worker, args=(spk_id, input_audio_path, voice_transform, unique_id, output_queue))
+        p.start()
+    
+    # Wait for the process to finish and get the result
+        p.join()
+        output_path = output_queue.get()
+        #output_path = convert_voice(spk_id, vocal_path, voice_transform,unique_id)
         output_path1= combine_vocal_and_inst(output_path,inst,unique_id)
         
         processed_audio_storage[unique_id] = output_path1
@@ -193,7 +204,7 @@ def api_convert_voice():
         
         print(output_path1)
         
-        created_files.extend([vocal_path, inst, output_path])
+        #created_files.extend([vocal_path, inst, output_path])
         return jsonify({"message": "File processed successfully", "audio_id": unique_id}), 200
     finally:
         request_semaphore.release()
@@ -225,7 +236,15 @@ def get_processed_audio(audio_id):
         return send_file(file_path, as_attachment=True)
     return jsonify({"error": "File not found."}), 404
 
-
+def worker(spk_id, input_audio_path, voice_transform, unique_id, output_queue):
+    """
+    Worker function to be executed in a separate process.
+    """
+    # Call the convert_voice function
+    output_audio_path = convert_voice(spk_id, input_audio_path, voice_transform, unique_id)
+    
+    # Put the result in the queue to be retrieved by the main process
+    output_queue.put(output_audio_path)
 def convert_voice(spk_id, input_audio_path, voice_transform,unique_id):
     get_vc(spk_id,0.5)
     print("*****before makinf call to vc ", unique_id)