experiments
/
STT
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Don't throw on mono audio any more since everything should work?

This commit is contained in:
CatalinVoss 2021-03-16 09:29:47 -07:00 committed by Reuben Morais
parent 900a01305c
commit be5f9627da
1 changed files with 6 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
training/coqui_stt_training/util/audio.py
View File

@ -579,17 +579,18 @@ def get_dtype(audio_format):
def pcm_to_np(pcm_data, audio_format=DEFAULT_FORMAT): def pcm_to_np(pcm_data, audio_format=DEFAULT_FORMAT):
if audio_format.channels != 1: # Handles both mono and stero audio
raise ValueError('Mono-channel audio required')
dtype = get_dtype(audio_format) dtype = get_dtype(audio_format)
samples = np.frombuffer(pcm_data, dtype=dtype) samples = np.frombuffer(pcm_data, dtype=dtype)
samples = samples.astype(np.float32) / np.iinfo(dtype).max samples = samples.astype(np.float32) / np.iinfo(dtype).max
if audio_format.channels == 1:
return np.expand_dims(samples, axis=1) return np.expand_dims(samples, axis=1)
else:
return samples
def np_to_pcm(np_data, audio_format=DEFAULT_FORMAT): def np_to_pcm(np_data, audio_format=DEFAULT_FORMAT):
if audio_format.channels != 1:
raise ValueError('Mono-channel audio required')
dtype = get_dtype(audio_format) dtype = get_dtype(audio_format)
np_data = np_data.squeeze() np_data = np_data.squeeze()
np_data = np_data * np.iinfo(dtype).max np_data = np_data * np.iinfo(dtype).max