Upload folder using huggingface_hub

app.py

@@ -8,6 +8,7 @@ from pathlib import Path
 from typing import List, Optional, Tuple
 
 import gradio as gr
+import librosa
 import numpy as np
 import soundfile as sf
 import torch
@@ -198,7 +199,30 @@ def resample_waveform(
 
 def load_audio_file(file_path: str, log: bool = True) -> Tuple[torch.Tensor, int]:
     log_progress(f"Loading audio: {Path(file_path).name}", enabled=log)
-    waveform, sample_rate = torchaudio.load(file_path)
+    waveform = None
+    sample_rate = None
+
+    try:
+        waveform, sample_rate = torchaudio.load(file_path)
+        waveform = ensure_mono(waveform)
+    except Exception as exc:
+        log_progress(f"torchaudio load failed: {exc}", 2, enabled=log)
+
+    if waveform is None or sample_rate is None:
+        try:
+            data, sample_rate = sf.read(
+                file_path, always_2d=True, dtype="float32"
+            )
+            waveform = torch.from_numpy(data.T)
+        except Exception as exc:
+            log_progress(f"soundfile load failed: {exc}", 2, enabled=log)
+            data, sample_rate = librosa.load(
+                file_path, sr=None, mono=False, dtype=np.float32
+            )
+            if data.ndim == 1:
+                data = data[None, :]
+            waveform = torch.from_numpy(data)
+
     waveform = ensure_mono(waveform)
     if sample_rate != DEFAULT_SAMPLE_RATE:
         log_progress(