import gradio as gr
from sbv2_bindings import TTSModel
from scipy.io import wavfile
import io


model = TTSModel.from_path("models/deberta.onnx", "models/tokenizer.json")
model.load_sbv2file_from_path("amitaro", "models/amitaro.sbv2")


def generate(text):
    style_vector = model.get_style_vector("amitaro", 0, 1.0)
    wav_data = model.synthesize(text, "amitaro", style_vector, 0.0, 0.5)
    byte_io = io.BytesIO(wav_data)
    sr, data = wavfile.read(byte_io)
    return sr, data


gradio_app = gr.Interface(
    fn=generate,
    inputs=["text"],
    outputs=[gr.Audio()],
)

if __name__ == "__main__":
    gradio_app.launch()