import torch
import os
rootdir=os.getcwd()
os.environ['TTS_HOME']=rootdir

from TTS.api import TTS
from dotenv import load_dotenv
load_dotenv()

print("源码部署需要先运行该文件，以便同意coqou-ai协议，当弹出协议时，请输入 y \n同时需要连接墙外下载或更新模型，请在 .env 中 HTTP_PROXY=设置代理地址")

def updatecache():
    # 禁止更新，避免无代理时报错
    file=os.path.join(rootdir,'tts_cache/cache')
    if file:
        import json,time
        j=json.load(open(file,'r',encoding='utf-8'))
        for i,it in enumerate(j):
            if "time" in it and "fn" in it:
                cache_file=os.path.join(rootdir,f'tts_cache/{it["fn"]}')
                if os.path.exists(cache_file) and os.path.getsize(cache_file)>17000000:
                    it['time']=time.time()
                    j[i]=it
        json.dump(j,open(file,'w',encoding='utf-8'))

updatecache()


device = "cuda" if torch.cuda.is_available() else "cpu"



#ttsv2 = TTS("tts_models/multilingual/multi-dataset/xtts_v2").to(device)

tts = TTS(model_name='voice_conversion_models/multilingual/vctk/freevc24').to(device)

# test
#tts.tts_to_file(text='我是中国人，你呢我的宝贝。今天天气看起来很不错啊', speaker_wav='./cn1.wav',language='zh', file_path='hafalse2.wav', speed=2.0,split_sentences=False)

#tts.tts_to_file(text='我是中国人，你呢我的宝贝。今天天气看起来很不错啊', speaker_wav='./cn1.wav',language='zh', file_path='hafalse0.2.wav', speed=0.2,split_sentences=False)

#target_wav is voice file 
# tts.voice_conversion_to_file(source_wav="./cn1.wav", target_wav="./sx1.wav", file_path="./out.wav")