app.py

import streamlit as st
import os
from audio_recorder_streamlit import audio_recorder
from streamlit_float import *
from openai import OpenAI



#set title
st.title(':rainbow[Audio to Text Converter ]🤖')

#set info container for information and fetch API key 
with st.popover(label=":information_source:",help="Enter your API key"):
     api_key = st.text_input("Enter Open AI key?",type="password")
     multi = '''This is a demo Audio to Text app,
 Supported input languages are :orange[English, Hindi, Marathi, etc.]'''

 
     
     st.markdown(multi)

     st.image("https://raw.githubusercontent.com/openai/whisper/main/language-breakdown.svg",output_format="auto")

     st.markdown("**Note:** For more details details on OpenAI-WHISPER, you can refer documentation(https://github.com/openai/whisper?tab=readme-ov-file)].")
     st.markdown("Created with  \
                :heart: by [Predikly Team](XXXXXXXXXXXXXXXXXXXXXXXXXX)")


client = OpenAI(api_key=api_key)

def speech_to_text(audio_data):
    #convert audio to text
    
    with open(audio_data, "rb") as audio_file:
        
        translation = client.audio.translations.create(
            model="whisper-1",
            file=audio_file
            )   
    return translation.text

#float feature intialization
float_init()

def initialize_session_state():
    
    if "message" not in st.session_state:
        st.session_state.message = [
            {"role":"Dipti","content":"How I can help you?"}
            ]
    
initialize_session_state()    

#create footer container
footer_container = st.container()
if api_key:
    with footer_container:
        audio_bytes = audio_recorder(
            energy_threshold=(-10.0, 10.0),
            pause_threshold=10.0,
            text="Click to record",
            recording_color="#e8b62c",
            neutral_color="#6aa36f",
            icon_name="microphone",
            icon_size="2x",
            key="audio_recorder")
else:
    st.warning("Please enter your API key using above ℹ icon")  
    audio_bytes=None #assigning blank value to audio_bytes variable to avoid error 


    
    
    

for message in st.session_state.message:
    with st.chat_message(message["role"],avatar=":material/robot_2:"):
        st.write(message["content"])   

if audio_bytes:
    #write audio byte to audio file
    with st.spinner("Converting..."):
        webm_file_path="audio.wav"
        with open(webm_file_path,"wb") as f:
            f.write(audio_bytes) 

        transcript = speech_to_text(webm_file_path)
        if transcript:
            st.session_state.message.append({"role":"user","content":transcript})
            with st.chat_message("user",avatar=":material/record_voice_over:"):
                st.write(transcript)
            os.remove(webm_file_path)