lora / 不知道 /回收站 /ext_voice.py
lintonxue00's picture
Upload ext_voice.py
61f1088
raw
history blame
7.1 kB
'''
启用tecent翻译 可以在YAML 中填入下面的参数
ng_voice_translate_on : True
tencentcloud_common_region : "ap-shanghai"
tencentcloud_common_secretid : "xxxxx"
tencentcloud_common_secretkey : "xxxxx"
ng_voice_tar : 'ja'
'''
from .Extension import Extension
import urllib, requests, uuid, os, base64
from aiohttp import request
from binascii import b2a_base64
from hashlib import sha1
from hmac import new
from random import randint
from sys import maxsize, version_info
from time import time
from nonebot import get_driver
from aiohttp import request
from loguru import logger
from nonebot.exception import ActionFailed
import asyncio
import requests
import base64
import uuid
import re
try:
from ujson import loads as loadJsonS
except:
from json import loads as loadJsonS
# 拓展的配置信息,用于ai理解拓展的功能 *必填*
ext_config:dict = {
"name": "voice", # 拓展名称,用于标识拓展
"arguments": {
'sentence': 'str', # 需要转换的文本
},
# 拓展的描述信息,用于提示ai理解拓展的功能 *必填* 尽量简短 使用英文更节省token
"description": "Send a voice sentence. (usage in response: /#voice&你好#/)",
# 参考词,用于上下文参考使用,为空则每次都会被参考(消耗token)
"refer_word": [],
# 作者信息
"author": "KroMiose",
# 版本
"version": "0.0.2",
# 拓展简介
"intro": "发送语音消息(支持翻译)",
}
# 自定义扩展
class CustomExtension(Extension):
async def call(self, arg_dict: dict, ctx_data: dict) -> dict:
""" 当拓展被调用时执行的函数 *由拓展自行实现*
参数:
arg_dict: dict, 由ai解析的参数字典 {参数名: 参数值}
"""
custom_config:dict = self.get_custom_config() # 获取yaml中的配置信息
ng_voice_translate_on = custom_config.get('ng_voice_translate_on', False) # 是否启用翻译
tencentcloud_common_region = custom_config.get('tencentcloud_common_region', "ap-shanghai") # 腾讯翻译-地区
tencentcloud_common_secretid = custom_config.get('tencentcloud_common_secretid',"xxxxx") # 腾讯翻译-密钥id
tencentcloud_common_secretkey = custom_config.get('tencentcloud_common_secretkey', "xxxxx") # 腾讯翻译-密钥
ng_voice_tar = custom_config.get('g_voice_tar', 'ja') # 翻译目标语言
is_base64 = custom_config.get('is_base64', False) # 是否使用base64编码
voice_path = 'voice_cache/'
# 创建缓存文件夹
if not os.path.exists(voice_path):
os.mkdir(voice_path)
# 获取参数
raw_text = arg_dict.get('sentence', None)
# 从这里开始翻译
# 腾讯翻译-签名
config = get_driver().config
async def getReqSign(params: dict) -> str:
common = {
"Action": "TextTranslate",
"Region": f"{tencentcloud_common_region}",
"Timestamp": int(time()),
"Nonce": randint(1, maxsize),
"SecretId": f"{tencentcloud_common_secretid}",
"Version": "2018-03-21",
}
params.update(common)
sign_str = "POSTtmt.tencentcloudapi.com/?"
sign_str += "&".join("%s=%s" % (k, params[k]) for k in sorted(params))
secret_key = tencentcloud_common_secretkey
if version_info[0] > 2:
sign_str = bytes(sign_str, "utf-8")
secret_key = bytes(secret_key, "utf-8")
hashed = new(secret_key, sign_str, sha1)
signature = b2a_base64(hashed.digest())[:-1]
if version_info[0] > 2:
signature = signature.decode()
return signature
async def q_translate(message) -> str:
_source_text = message
_source = "auto"
_target = ng_voice_tar
try:
endpoint = "https://tmt.tencentcloudapi.com"
params = {
"Source": _source,
"SourceText": _source_text,
"Target": _target,
"ProjectId": 0,
}
params["Signature"] = await getReqSign(params)
# 加上超时参数
async with request("POST", endpoint, data=params) as resp:
data = loadJsonS(await asyncio.wait_for(resp.read(), timeout=30))["Response"]
message = data["TargetText"]
except ActionFailed as e:
logger.warning(
f"ActionFailed {e.info['retcode']} {e.info['msg'].lower()} {e.info['wording']}"
)
except TimeoutError as e:
logger.warning(
f"TimeoutError {e}"
)
return message
# 到这里翻译函数结束
if ng_voice_translate_on == True :
t_result = await q_translate(raw_text)
else:
t_result = raw_text
text = t_result + '~' # 加上一个字符,避免合成语音丢失结尾
text = urllib.parse.quote(text) # url编码
# ! moegoe.azurewebsites.net 是一个公开的语音合成api,非本人搭建,请勿滥用 (tip by KroMiose)
# 该api只支持日语,如果传入其它语言,会导致合成结果不可预知
# 如果你开启了翻译就没事了
url = f"https://lintonxue00-vits-umamusume-voice-synthesizer.hf.space/api/tts?text={text}&speaker_id=0"
# todo: 如果需要使用本地的语音合成api,请取消注释下面的代码,并自行改为您的api地址,将填入合成文本的地方改为{text}
# url = f"http://127.0.0.1:23211/to_voice?text={text}"
# 去除base64字符串中的非法字符
def call(self, args, context):
url = args['url']
text = args.get('text')
is_base64 = args.get('is_base64', False)
raw_text = text
# 下载语音文件
r = requests.get(url)
# 对base64字符串进行处理
if is_base64:
pattern = re.compile('[^a-zA-Z0-9+/=]')
content = pattern.sub('', r.content.decode())
audio_data = base64.b64decode(content + b'=' * (-len(content) % 4), validate=True)
else:
audio_data = r.content
# 写入音频文件
file_name = f"{voice_path}{uuid.uuid1()}.ogg"
with open(file_name, "wb") as f:
f.write(audio_data)
local_url = f"file:///{os.path.abspath(file_name)}"
if text is not None:
return {
'voice': local_url, # 语音url
'text': f"[语音消息] {raw_text}", # 文本
}
return {}
def __init__(self, custom_config: dict):
super().__init__(ext_config.copy(), custom_config)