Spaces:
Sleeping
Sleeping
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 |