app.py

import streamlit as st
import torch
import torchaudio
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor

# Load the model and tokenizer
processor = Wav2Vec2Processor.from_pretrained("facebook/wav2vec2-base-960h")
model = Wav2Vec2ForCTC.from_pretrained("facebook/wav2vec2-base-960h")

# Define a function to transcribe audio
def transcribe(audio_file):
    audio, sample_rate = torchaudio.load(audio_file)
    input_values = processor(audio, sampling_rate=sample_rate, return_tensors="pt").input_values
    logits = model(input_values).logits
    predicted_ids = torch.argmax(logits, dim=-1)
    transcription = processor.decode(predicted_ids[0])
    return transcription

# Set up the Streamlit app
st.title("Speech Recognition with Wav2Vec2")
audio_file = st.file_uploader("Upload an audio file", type=["mp3", "wav"])

if audio_file is not None:
    st.audio(audio_file, format="audio/wav")
    transcript = transcribe(audio_file)
    st.write("Transcription: ", transcript)