modules/tts_frontend.py

import os

import soundfile as sf
from espnet_onnx import Text2Speech
from pydantic import BaseModel


class TextInput(BaseModel):
    text: str = ""


class TTSModel:
    def __init__(self):
        self.model = None
        self.model_path = "models/tts"
        self.output_path = "template/sample.wav"

    def load_model(self, model_path):
        self.model = Text2Speech(model_dir=f"{self.model_path}/{model_path}")

    def generate(self, text_input: TextInput):
        if self.model is None:
            raise RuntimeError("Model is not loaded.")

        if os.path.exists(self.output_path):
            os.remove(self.output_path)

        audio = self.model(text_input.text)["wav"]
        sf.write(self.output_path, audio, 16000)
        return