rcell commited on
Commit
99185bb
1 Parent(s): f5d0e3c
Files changed (1) hide show
  1. app.py +7 -6
app.py CHANGED
@@ -9,7 +9,7 @@ numba_logger.setLevel(logging.WARNING)
9
 
10
  import librosa
11
  import torch
12
- import torchcrepe
13
  import commons
14
  import utils
15
  from models import SynthesizerTrn
@@ -19,14 +19,15 @@ def resize2d(source, target_len):
19
  source[source<0.001] = np.nan
20
  target = np.interp(np.arange(0, len(source)*target_len, len(source))/ target_len, np.arange(0, len(source)), source)
21
  return np.nan_to_num(target)
22
-
23
  def convert_wav_22050_to_f0(audio):
24
- tmp = torchcrepe.predict(audio=audio, fmin=50, fmax=550,
25
- sample_rate=22050, model='full',
26
- batch_size=2048).numpy()[0]
 
27
  f0 = np.zeros_like(tmp)
28
- f0[tmp > 0] = tmp[tmp > 0]
29
  return f0
 
30
  def get_text(text, hps):
31
  text_norm = text_to_sequence(text, hps.data.text_cleaners)
32
  if hps.data.add_blank:
 
9
 
10
  import librosa
11
  import torch
12
+
13
  import commons
14
  import utils
15
  from models import SynthesizerTrn
 
19
  source[source<0.001] = np.nan
20
  target = np.interp(np.arange(0, len(source)*target_len, len(source))/ target_len, np.arange(0, len(source)), source)
21
  return np.nan_to_num(target)
 
22
  def convert_wav_22050_to_f0(audio):
23
+ tmp = librosa.pyin(audio,
24
+ fmin=librosa.note_to_hz('C0'),
25
+ fmax=librosa.note_to_hz('C7'),
26
+ frame_length=1780)[0]
27
  f0 = np.zeros_like(tmp)
28
+ f0[tmp>0] = tmp[tmp>0]
29
  return f0
30
+
31
  def get_text(text, hps):
32
  text_norm = text_to_sequence(text, hps.data.text_cleaners)
33
  if hps.data.add_blank: