wip

app.py

@@ -8,7 +8,6 @@ import markdown
 import matplotlib.pyplot as plt
 import numpy as np
 import torch
-import torchaudio as ta
 from loguru import logger
 
 from df import config
@@ -72,16 +71,15 @@ def mix_and_denoise(speech_rec, speech_upl, noise, snr):
         speech_file = speech_rec
         sp_kwargs = {"frame_offset": 4800}
     meta = None
-    try:
-        print(ta.info(speech_file))
-        speech, meta = load_audio(speech_file, sr, **sp_kwargs)
-    except RuntimeError as e:
-        if meta is not None:
-            print(meta)
-        raise
+    for f in ("wav", "mp3", "flac", "vorbis", "opus"):
+        print(f)
         # Chrome saves files as opus now.
-        sp_kwargs["format"] = "opus"
-        speech, meta = load_audio(speech_file, sr, **sp_kwargs)
+        sp_kwargs["format"] = f
+        try:
+            speech, meta = load_audio(speech_file, sr, **sp_kwargs)
+        except RuntimeError as e:
+            if meta is not None:
+                print(meta)
 
     logger.info(f"Loaded speech with shape {speech.shape}")
     noise, _ = load_audio(noise, sr)