Spaces:
Running
Running
Fix audio pre-processing
Browse files
app.py
CHANGED
@@ -29,13 +29,15 @@ def greet(image, audio):
|
|
29 |
|
30 |
audio_file = torch.from_numpy(audio)
|
31 |
|
|
|
|
|
|
|
|
|
|
|
32 |
if desired_sample_rate != sample_rate:
|
33 |
audio_file = torchaudio.functional.resample(audio_file, sample_rate, desired_sample_rate)
|
34 |
|
35 |
-
|
36 |
-
audio_file = torch.concat([audio_file[0], audio_file[1]], dim=0) # Stereo -> mono (x2 duration)
|
37 |
-
|
38 |
-
audio_file.squeeze(0)
|
39 |
|
40 |
if audio_file.shape[0] > (desired_sample_rate * set_length):
|
41 |
audio_file = audio_file[:desired_sample_rate * set_length]
|
|
|
29 |
|
30 |
audio_file = torch.from_numpy(audio)
|
31 |
|
32 |
+
if len(audio_file.shape) == 2:
|
33 |
+
audio_file = torch.concat([audio_file[:, 0:1], audio_file[:, 1:2]], dim=0).T # Stereo -> mono (x2 duration)
|
34 |
+
else:
|
35 |
+
audio_file = audio_file.unsqueeze(0)
|
36 |
+
|
37 |
if desired_sample_rate != sample_rate:
|
38 |
audio_file = torchaudio.functional.resample(audio_file, sample_rate, desired_sample_rate)
|
39 |
|
40 |
+
audio_file = audio_file.squeeze(0)
|
|
|
|
|
|
|
41 |
|
42 |
if audio_file.shape[0] > (desired_sample_rate * set_length):
|
43 |
audio_file = audio_file[:desired_sample_rate * set_length]
|