app.py

import torch
import torchaudio
from transformers import Wav2Vec2Processor, Wav2Vec2ForCTC, Wav2Vec2CTCTokenizer

# Load the Arabic-specific processor and model
model_name = "Zaid/wav2vec2-large-xlsr-53-arabic-egyptian"
tokenizer = Wav2Vec2CTCTokenizer.from_pretrained(model_name)
processor = Wav2Vec2Processor.from_pretrained(model_name, tokenizer=tokenizer)
model = Wav2Vec2ForCTC.from_pretrained(model_name)

def transcribe(audio_file):
    try:
        # Load the audio file
        print("Loading audio file...")
        audio_input, sr = torchaudio.load(audio_file)
        print(f"Audio loaded: {audio_input.shape}, Sample rate: {sr}")

        # Resample if needed
        if sr != 16000:
            print("Resampling audio...")
            resampler = torchaudio.transforms.Resample(orig_freq=sr, new_freq=16000)
            audio_input = resampler(audio_input)
            sr = 16000
        
        print(f"Audio shape after resampling: {audio_input.shape}, Sample rate: {sr}")

        # Convert tensor to numpy array
        audio_input = audio_input[0].numpy()

        # Process audio input
        print("Processing audio input...")
        input_values = processor(audio_input, return_tensors="pt", sampling_rate=sr).input_values

        # Run model inference
        print("Running model inference...")
        with torch.no_grad():
            logits = model(input_values).logits

        # Decode transcription
        print("Decoding transcription...")
        predicted_ids = torch.argmax(logits, dim=-1)
        transcription = processor.batch_decode(predicted_ids, skip_special_tokens=True, clean_up_tokenization_spaces=True)
        return transcription[0]
    except Exception as e:
        print(f"An error occurred: {e}")
        return None

# Transcribe the audio file
transcription = transcribe("sidiali.wav")
if transcription:
    print(transcription.encode('utf-8').decode('utf-8'))
    # Save the transcription to a file
    with open("transcription.txt", "w", encoding="utf-8") as f:
        f.write(transcription)
    print("Transcription saved to transcription.txt")