Add application file

app.py

@@ -21,7 +21,6 @@ import logging
 import time
 import traceback
 import numpy as np
-import yt_dlp
 from pathlib import Path
 from huggingface_hub import hf_hub_download
 from typing import Dict, Tuple
@@ -43,6 +42,35 @@ STEM_NAMING = {
     "Bass": "Bassless",
 }
 
+
+def convert_to_stereo_and_wav(audio_path: Path) -> Path:
+    # loading takes time since resampling at 44100 Hz
+    wave, sr = librosa.load(str(audio_path), mono=False, sr=44100)
+
+    # check if mono
+    if type(wave[0]) != np.ndarray or audio_path.suffix != ".wav": # noqa
+        stereo_path = audio_path.with_name(audio_path.stem + "_stereo.wav")
+
+        command = shlex.split(
+            f'ffmpeg -y -loglevel error -i "{str(audio_path)}" -ac 2 -f wav "{str(stereo_path)}"'
+        )
+        sub_params = {
+            "stdout": subprocess.PIPE,
+            "stderr": subprocess.PIPE,
+            "creationflags": subprocess.CREATE_NO_WINDOW
+            if sys.platform == "win32"
+            else 0,
+        }
+        process_wav = subprocess.Popen(command, **sub_params)
+        output, errors = process_wav.communicate()
+        if process_wav.returncode != 0 or not stereo_path.exists():
+            raise Exception("Error processing audio to stereo wav")
+
+        return stereo_path
+    else:
+        return Path(audio_path)
+    
+
 class MDXModel:
     def __init__(
         self,