File size: 641 Bytes
597a3c5 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 |
# This is used to detect the spoken language in an audio file
# I wanted to abstract it to it's own file, just like vocal isolation & diarization
from speechbrain.pretrained import EncoderClassifier
language_identifier_model = None
def detect_language(file):
global language_identifier_model
if not language_identifier_model:
language_identifier_model = EncoderClassifier.from_hparams(source="speechbrain/lang-id-voxlingua107-ecapa", savedir="tmp") #, run_opts={"device":"cuda"})
signal = language_identifier_model.load_audio(file)
prediction = language_identifier_model.classify_batch(signal)
return prediction[3][0].split(' ')[1]
|