from TTSInferencing import TTSInferencing
import  streamlit as st
from speechbrain.inference.vocoders import HIFIGAN
import torchaudio


tts_model = TTSInferencing.from_hparams(source="./",
                                        hparams_file='./hyperparams.yaml',
                                        pymodule_file='./module_classes.py',
                                        savedir="/results/",
                                        )

hifi_gan = HIFIGAN.from_hparams(source="speechbrain/tts-hifigan-ljspeech", savedir='/pretrained_models/hifi-gan-ljspeech')

# text = ["Hello I am a girl", "How is your day going", "I hope you are doing well"]

# Input text
text = [st.text_input("Enter your text here")]

if text:
    mel_outputs = tts_model.encode_batch(text)
    waveforms = hifi_gan.decode_batch(mel_outputs)

    waveform =  waveforms[0].squeeze(1)

    # Display the audio widget to play the synthesized speech
    st.audio(waveform.numpy(), format="audio/wav")