import torch
from transformers import Wav2Vec2ForCTC, Wav2Vec2Tokenizer
import gradio as gr

# Inisialisasi model dan tokenizer
model_path = 'saved_model/model.pth'
model = Wav2Vec2ForCTC.from_pretrained(model_path)
tokenizer = Wav2Vec2Tokenizer.from_pretrained(model_path)

def text_to_speech(text):
    input_ids = tokenizer.encode(text, return_tensors="pt").squeeze(0)
    logits = model.generate(input_ids)
    audio = model.get_feature_vector(logits)
    return audio

def tts(text):
    audio = text_to_speech(text)
    return audio

inputs = gr.inputs.Textbox(label="Masukkan teks (Bahasa Jepang)")
outputs = gr.outputs.Audio(label="Audio")

title = "Text-to-Speech dengan Model VITS-SVC Gmodel"
description = "Demo TTS menggunakan model VITS-SVC Gmodel untuk bahasa Jepang."
gr.Interface(fn=tts, inputs=inputs, outputs=outputs, title=title, description=description).launch()