Views
No views yet
1import librosa
2
3processor = WhisperProcessor.from_pretrained('rverma251/sanskrit-whisper-asr')
4model = WhisperForConditionalGeneration.from_pretrained('rverma251/sanskrit-whisper-asr')
5
6model.generation_config.language = 'hindi'
7model.generation_config.task = 'transcribe'
8
9audio, sr = librosa.load('sanskrit_audio.wav', sr=16000)
10inputs = processor.feature_extractor(audio, sampling_rate=16000, return_tensors='pt')
11predicted_ids = model.generate(inputs.input_features)
12transcription = processor.batch_decode(predicted_ids, skip_special_tokens=True)
13print(transcription)