import gradio as gr
import os
import torchaudio
from fireredtts.fireredtts import FireRedTTS


gpt_path = 'https://huggingface.co/fireredteam/FireRedTTS/resolve/main/fireredtts_gpt.pt'
speaker_path = 'https://huggingface.co/fireredteam/FireRedTTS/resolve/main/fireredtts_speaker.bin'
decoder_path = 'https://huggingface.co/fireredteam/FireRedTTS/resolve/main/fireredtts_token2wav.pt'


tts = FireRedTTS(
    config_path="configs/config_24k.json",
    pretrained_path='pretrained_models',
)

def tts_inference(text, prompt_wav='examples/prompt_1.wav', lang='zh'):
    syn_audio = tts.synthesize(
        prompt_wav=prompt_wav,
        text=text,
        lang=lang,
    )
    return syn_audio.detach().cpu()


iface = gr.Interface(
    fn=tts_inference,
    inputs=[
        gr.Textbox(label="输入文本"),
        # gr.Dropdown(["en-US-Wav2Vec2-Kendra", "en-US-Wav2Vec2-John"], label="选择声音"),
        # gr.Slider(minimum=0.5, maximum=2.0, value=1.0, label="语速")
    ],
    outputs=gr.Audio(label="生成的语音"),
    title="TTS Demo", 
    description="Enter some text and listen to the generated speech."
)

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