Print file properties

app.py

@@ -1,16 +1,16 @@
+import math
 import tempfile
-from df.utils import resample
 
 import gradio
 import gradio.inputs
 import gradio.outputs
-import matplotlib.pyplot as plt
 import markdown
+import matplotlib.pyplot as plt
 import numpy as np
 import torch
 from df import config
-import math
 from df.enhance import enhance, init_df, load_audio, save_audio
+from df.utils import resample
 
 device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 model, df, _ = init_df()
@@ -60,14 +60,24 @@ def mix_and_denoise(speech_rec, speech_upl, noise, snr):
         noise = "samples/dkitchen.wav"
     sp_kwargs = {}
     if speech_rec is None or "none" in speech_rec:
-        speech = "samples/p232_013_clean.wav"
+        speech_file = "samples/p232_013_clean.wav"
         if speech_upl is not None and "none" not in speech_upl:
             print("using speech_upl")
-            speech = speech_upl
+            speech_file = speech_upl
     else:
-        speech = speech_rec
+        speech_file = speech_rec
         sp_kwargs = {"frame_offset": 4800}
-    speech, meta = load_audio(speech, sr, **sp_kwargs)
+    try:
+        speech, meta = load_audio(speech_file, sr, **sp_kwargs)
+    except RuntimeError as e:
+        print("Could not load audio:", e)
+        import os
+
+        print(os.path.getsize(speech_file))
+        print(os.path.getmtime(speech_file))
+        print(os.path.getctime(speech_file))
+        raise e
+
     print(f"Loaded speech with shape {speech.shape}")
     noise, _ = load_audio(noise, sr)
     if meta.sample_rate != sr: