Python script based on gradio library to create a quick website showcasing the ability to use audio to interact with ChatGPT

banglaspeakChatGPT.py

import config
from google.cloud import speech, texttospeech
import gradio as gr
import io
import openai
import os
from playsound import playsound

openai.api_key = config.OPENAI_API_KEY ## API Key from ChatGPT stored in a config.py file in same directory
os.environ["GOOGLE_APPLICATION_CREDENTIALS"] = "./cloud_creds.json" ## Path to service account credentials JSON from Google Cloud Platform

# The text that will be spoken using playsound
text_for_speech = ""

ROLE_MAPPING = {"assistant": "ChatGPT", "user": "আপনি"}
# Initial message provided to ChatGPT
messages = [
    {
        "role": "system",
        "content": "You are a helpful assistant. Respond to all input within 100 words.",
    }
]

# Create a string representation of the messages list that can be displayed
def convert_messages_to_text():
    global messages

    chat_transcript = ""
    for message in messages:
        if message["role"] != "system":
            chat_transcript += (
                ROLE_MAPPING[message["role"]] + ": " + message["content"] + "\n\n"
            )

    return chat_transcript

# Convert the text to audio file and then play it
def play_audio(audio):
    global text_for_speech

    client = texttospeech.TextToSpeechClient()
    synthesis_input = texttospeech.SynthesisInput(text=text_for_speech)

    voice = texttospeech.VoiceSelectionParams(
        language_code="bn-BD", ssml_gender=texttospeech.SsmlVoiceGender.FEMALE
    )

    audio_config = texttospeech.AudioConfig(
        audio_encoding=texttospeech.AudioEncoding.MP3
    )

    response = client.synthesize_speech(
        input=synthesis_input, voice=voice, audio_config=audio_config
    )

    with open("output.mp3", "wb") as out:
        out.write(response.audio_content)
        print('Audio content written to file "output.mp3"')

    playsound("./output.mp3")
    chat_transcript = convert_messages_to_text()

    return chat_transcript

# Receive response from ChatGPT and add it to the global messages list
def get_response_from_chatgpt(text):
    global text_for_speech, messages

    messages.append({"role": "user", "content": text})

    response = openai.ChatCompletion.create(model="gpt-3.5-turbo", messages=messages)

    system_message = response["choices"][0]["message"]

    messages.append(system_message)

    chat_transcript = convert_messages_to_text()

    text_for_speech = system_message["content"]

    return chat_transcript

# Converted recorded audio into text
def get_text_from_speech(audio_path):
    client = speech.SpeechClient()

    if not audio_path:
        return None

    with io.open(audio_path, "rb") as audio_file:
        content = audio_file.read()

    audio = speech.RecognitionAudio(content=content)

    config = speech.RecognitionConfig(
        encoding=speech.RecognitionConfig.AudioEncoding.LINEAR16,
        sample_rate_hertz=48000,
        language_code="bn-BD",
    )

    response = client.recognize(config=config, audio=audio)

    transcript = ""
    for result in response.results:
        transcript += result.alternatives[0].transcript

    return transcript


def transcribe(audio):

    transcript = get_text_from_speech(audio)
    if not transcript:
        return "অনুগ্রহ করে আপনার প্রশ্নটি আবার রেকর্ড করুন এবং সাবমিট বাটন ক্লিক করুন"
    chat_transcript = get_response_from_chatgpt(transcript)

    return chat_transcript


with gr.Blocks(title="বাংলায় ChatGPT", css="h1, p {text-align: center}") as demo:
    gr.Markdown(
        """
    # বাংলায় কথা বলুন এবং উত্তর পান ChatGPT থেকে!
    এই ওয়েবসাইট ব্যাবহার করে আপনি বাংলায় আপনার অডিও রেকর্ড করে ChatGPT কে পাঠাতে পারবেন এবং AI থেকে বাংলায় উত্তর পাবেন!
    """
    )
    audio = gr.Audio(
        source="microphone", type="filepath", label="আপনার প্রশ্ন রেকর্ড করুন"
    )
    submit_button = gr.Button("আপনার প্রশ্ন জমা দিন!")
    output = gr.Textbox(
        label="chatgpt থেকে উত্তর:",
        placeholder="উত্তর এখানে দেখানো হবে",
    )
    audio_play = gr.Button("উত্তরের অডিও শুনতে ক্লিক করুন!")

    submit_button.click(fn=transcribe, inputs=audio, outputs=output)
    audio_play.click(fn=play_audio, inputs=audio, outputs=output)

demo.launch()