import streamlit as st
from transformers import pipeline
from gtts import gTTS
from io import BytesIO
import base64

# Define a function to generate text-to-speech audio
def generate_tts(text, voice_type, language):
    # Hugging Face pipeline for text-to-speech in English
    if language == "English":
        model_id = "facebook/fastspeech2-en-ljspeech"
        tts_pipeline = pipeline("text-to-speech", model=model_id)
        audio_data = tts_pipeline(text, return_tensors=True).audio["array"]
        return audio_data
    # Google Text-to-Speech (gTTS) for Urdu
    elif language == "Urdu":
        tts = gTTS(text=text, lang="ur")
        mp3_fp = BytesIO()
        tts.write_to_fp(mp3_fp)
        mp3_fp.seek(0)
        return mp3_fp.read()

# Configure Streamlit app
st.title("Multilingual Text-to-Speech Application")
st.write("Generate speech from text in American English and Urdu with different voice styles.")

# Input fields for text and options
text = st.text_area("Enter your text:", "Hello, welcome to the Text-to-Speech app!")
voice_type = st.selectbox(
    "Select voice style:", ["Adult", "Child", "Cartoon"]
)
language = st.selectbox("Select language:", ["English", "Urdu"])

if st.button("Generate Audio"):
    with st.spinner("Generating audio..."):
        try:
            audio_data = generate_tts(text, voice_type, language)
            if language == "English":
                # Provide the audio as a downloadable file
                st.audio(audio_data, format="audio/wav")
            elif language == "Urdu":
                b64 = base64.b64encode(audio_data).decode()
                href = f'<a href="data:audio/mpeg;base64,{b64}" download="output.mp3">Download Urdu Audio</a>'
                st.markdown(href, unsafe_allow_html=True)
        except Exception as e:
            st.error(f"Error: {e}")

st.caption("Powered by Hugging Face Transformers and Google TTS")