app.py

import subprocess
subprocess.run(["pip", "install", "gradio", "--upgrade"])
subprocess.run(["pip", "install", "transformers"])
subprocess.run(["pip", "install", "torchaudio", "--upgrade"])

import gradio as gr
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
import torchaudio

# Load model and processor
processor = Wav2Vec2Processor.from_pretrained("jonatasgrosman/wav2vec2-large-xlsr-53-italian")
model = Wav2Vec2ForCTC.from_pretrained("jonatasgrosman/wav2vec2-large-xlsr-53-italian")

# Function to perform ASR on audio data
def transcribe_audio(audio_data):
    # Convert audio data to mono and normalize
    audio_data = torchaudio.transforms.Mono()(audio_data)
    audio_data = torchaudio.functional.gain(audio_data, gain_db=5.0)

    # Resample if needed (Wav2Vec2 model requires 16 kHz sampling rate)
    if audio_data[1] != 16000:
        audio_data = torchaudio.transforms.Resample(audio_data[1], 16000)(audio_data[0])

    # Apply custom preprocessing to the audio data if needed
    input_values = processor(audio_data[0].numpy(), return_tensors="pt").input_values

    # Perform ASR
    with torch.no_grad():
        logits = model(input_values).logits

    # Decode the output
    predicted_ids = torch.argmax(logits, dim=-1)
    transcription = processor.batch_decode(predicted_ids)

    return transcription[0]

# Create Gradio interface
audio_input = gr.Audio()
gr.Interface(fn=transcribe_audio, inputs=audio_input, outputs="text").launch()