Views
No views yet
ٱ et ٰ).1from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
2import torch
3import librosa
4
5model_id = "VOTRE_NOM_UTILISATEUR/wav2vec2-large-xlsr-53-arabic-quran"
6
7processor = Wav2Vec2Processor.from_pretrained(model_id)
8model = Wav2Vec2ForCTC.from_pretrained(model_id)
9
10# Chargement audio (16kHz obligatoire)
11audio, sr = librosa.load("verset.mp3", sr=16000)
12
13inputs = processor(audio, sampling_rate=16000, return_tensors="pt", padding=True)
14
15with torch.no_grad():
16 logits = model(inputs.input_values).logits
17
18predicted_ids = torch.argmax(logits, dim=-1)
19transcription = processor.batch_decode(predicted_ids, skip_special_tokens=True)[0]
20
21print(transcription)