app.py

import streamlit as st 
import soundfile as sf
import librosa
from transformers import HubertForCTC, Wav2Vec2Processor , pipeline , Wav2Vec2ForCTC , Wav2Vec2Tokenizer
import torch
import spacy 
from spacy import displacy

st.title('Audio-to-Text')

audio_file = st.file_uploader('Upload Audio' , type=['wav' , 'mp3','m4a'])

if st.button('Trascribe Audio'):
    if audio_file is not None:
        processor = Wav2Vec2Processor.from_pretrained("facebook/wav2vec2-base-960h")
        model = HubertForCTC.from_pretrained("facebook/hubert-large-ls960-ft")  
        speech, rate = librosa.load(audio_file, sr=16000)
        input_values = processor(speech, return_tensors="pt", padding="longest", sampling_rate=rate).input_values
        logits = model(input_values).logits
        predicted_ids = torch.argmax(logits, dim=-1)
        text = processor.batch_decode(predicted_ids)
        st.write(text) 
    else:
        st.error('please upload the audio file')



if st.button('Summarize'):
        processor = Wav2Vec2Processor.from_pretrained("facebook/wav2vec2-base-960h")
        model = HubertForCTC.from_pretrained("facebook/hubert-large-ls960-ft")  
        speech, rate = librosa.load(audio_file, sr=16000)
        input_values = processor(speech, return_tensors="pt", padding="longest", sampling_rate=rate).input_values
        logits = model(input_values).logits
        predicted_ids = torch.argmax(logits, dim=-1)
        text = processor.batch_decode(predicted_ids)
        summarize = pipeline("summarization")
        st.write(summarize(text))

if st.button('sentiment-analysis'):
        processor = Wav2Vec2Processor.from_pretrained("facebook/wav2vec2-base-960h")
        model = HubertForCTC.from_pretrained("facebook/hubert-large-ls960-ft")  
        speech, rate = librosa.load(audio_file, sr=16000)
        input_values = processor(speech, return_tensors="pt", padding="longest", sampling_rate=rate).input_values
        logits = model(input_values).logits
        predicted_ids = torch.argmax(logits, dim=-1)
        text = processor.batch_decode(predicted_ids)
        nlp_sa = pipeline("sentiment-analysis")
        st.write(nlp_sa(text))

if st.button('Name'):
        processor = Wav2Vec2Processor.from_pretrained("facebook/wav2vec2-base-960h")
        model = HubertForCTC.from_pretrained("facebook/hubert-large-ls960-ft")  
        speech, rate = librosa.load(audio_file, sr=16000)
        input_values = processor(speech, return_tensors="pt", padding="longest", sampling_rate=rate).input_values
        logits = model(input_values).logits
        predicted_ids = torch.argmax(logits, dim=-1)
        text = processor.batch_decode(predicted_ids)
        str  = ''.join(text)
        trf = spacy.load('en_core_web_trf')
        doc=trf(str)
        print(displacy.render(doc,style='ent'))