Spaces:
Sleeping
Sleeping
File size: 815 Bytes
c6a12ae |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 |
import torch
import librosa
from transformers import WhisperProcessor, WhisperForConditionalGeneration
from config import whisper_model_name
whisper_processor = WhisperProcessor.from_pretrained(whisper_model_name)
whisper_model = WhisperForConditionalGeneration.from_pretrained(whisper_model_name)
def transcribe_speech(audiopath):
speech, rate = librosa.load(audiopath, sr=16000)
audio_input = whisper_processor(speech, return_tensors="pt", sampling_rate=16000)
with torch.no_grad():
generated_ids = whisper_model.generate(audio_input["input_features"])
transcription = whisper_processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
return transcription
def getAudioArray(audio_path):
speech, rate = librosa.load(audio_path, sr=16000)
return speech |