Merge pull request #1821 from Zhalslar/gsv-tts-selfhost

Feature: 新增 GPT_SoVIS 的 TTS 服务商
This commit is contained in:
Soulter
2025-06-21 23:58:07 +08:00
committed by GitHub
3 changed files with 307 additions and 0 deletions
+155
View File
@@ -850,6 +850,37 @@ CONFIG_METADATA_2 = {
"edge-tts-voice": "zh-CN-XiaoxiaoNeural",
"timeout": 20,
},
"GSV TTS(本地加载)": {
"id": "gsv_tts",
"enable": False,
"type": "gsv_tts_selfhost",
"provider_type": "text_to_speech",
"api_base": "http://127.0.0.1:9880",
"gpt_weights_path": "",
"sovits_weights_path": "",
"timeout": 60,
"gsv_default_parms": {
"gsv_ref_audio_path": "",
"gsv_prompt_text": "",
"gsv_prompt_lang": "zh",
"gsv_aux_ref_audio_paths": "",
"gsv_text_lang": "zh",
"gsv_top_k": 5,
"gsv_top_p": 1.0,
"gsv_temperature": 1.0,
"gsv_text_split_method": "cut3",
"gsv_batch_size": 1,
"gsv_batch_threshold": 0.75,
"gsv_split_bucket": True,
"gsv_speed_factor": 1,
"gsv_fragment_interval": 0.3,
"gsv_streaming_mode": False,
"gsv_seed": -1,
"gsv_parallel_infer": True,
"gsv_repetition_penalty": 1.35,
"gsv_media_type": "wav",
},
},
"GSVI TTS(API)": {
"id": "gsvi_tts",
"type": "gsvi_tts_api",
@@ -951,6 +982,130 @@ CONFIG_METADATA_2 = {
},
},
"items": {
"gpt_weights_path": {
"description": "GPT模型文件路径",
"type": "string",
"hint": "即“.ckpt”后缀的文件,请使用绝对路径,路径两端不要带双引号,不填则默认用GPT_SoVITS内置的SoVITS模型(建议直接在GPT_SoVITS中改默认模型)",
"obvious_hint": True,
},
"sovits_weights_path": {
"description": "SoVITS模型文件路径",
"type": "string",
"hint": "即“.pth”后缀的文件,请使用绝对路径,路径两端不要带双引号,不填则默认用GPT_SoVITS内置的SoVITS模型(建议直接在GPT_SoVITS中改默认模型)",
"obvious_hint": True,
},
"gsv_default_parms": {
"description": "GPT_SoVITS默认参数",
"hint": "参考音频文件路径、参考音频文本必填,其他参数根据个人爱好自行填写",
"type": "object",
"items": {
"gsv_ref_audio_path": {
"description": "参考音频文件路径",
"type": "string",
"hint": "必填!请使用绝对路径!路径两端不要带双引号!",
"obvious_hint": True,
},
"gsv_prompt_text": {
"description": "参考音频文本",
"type": "string",
"hint": "必填!请填写参考音频讲述的文本",
"obvious_hint": True,
},
"gsv_prompt_lang": {
"description": "参考音频文本语言",
"type": "string",
"hint": "请填写参考音频讲述的文本的语言,默认为中文",
},
"gsv_aux_ref_audio_paths": {
"description": "辅助参考音频文件路径",
"type": "string",
"hint": "辅助参考音频文件,可不填",
},
"gsv_text_lang": {
"description": "文本语言",
"type": "string",
"hint": "默认为中文",
},
"gsv_top_k": {
"description": "生成语音的多样性",
"type": "int",
"hint": "",
},
"gsv_top_p": {
"description": "核采样的阈值",
"type": "float",
"hint": "",
},
"gsv_temperature": {
"description": "生成语音的随机性",
"type": "float",
"hint": "",
},
"gsv_text_split_method": {
"description": "切分文本的方法",
"type": "string",
"hint": "可选值: `cut0`:不切分 `cut1`:四句一切 `cut2`50字一切 `cut3`:按中文句号切 `cut4`:按英文句号切 `cut5`:按标点符号切",
"options": [
"cut0",
"cut1",
"cut2",
"cut3",
"cut4",
"cut5",
],
},
"gsv_batch_size": {
"description": "批处理大小",
"type": "int",
"hint": "",
},
"gsv_batch_threshold": {
"description": "批处理阈值",
"type": "float",
"hint": "",
},
"gsv_split_bucket": {
"description": "将文本分割成桶以便并行处理",
"type": "bool",
"hint": "",
},
"gsv_speed_factor": {
"description": "语音播放速度",
"type": "float",
"hint": "1为原始语速",
},
"gsv_fragment_interval": {
"description": "语音片段之间的间隔时间",
"type": "float",
"hint": "",
},
"gsv_streaming_mode": {
"description": "启用流模式",
"type": "bool",
"hint": "",
},
"gsv_seed": {
"description": "随机种子",
"type": "int",
"hint": "用于结果的可重复性",
},
"gsv_parallel_infer": {
"description": "并行执行推理",
"type": "bool",
"hint": "",
},
"gsv_repetition_penalty": {
"description": "重复惩罚因子",
"type": "float",
"hint": "",
},
"gsv_media_type": {
"description": "输出媒体的类型",
"type": "string",
"hint": "建议用wav",
},
},
},
"embedding_dimensions": {
"description": "嵌入维度",
"type": "int",
+4
View File
@@ -225,6 +225,10 @@ class ProviderManager:
from .sources.edge_tts_source import (
ProviderEdgeTTS as ProviderEdgeTTS,
)
case "gsv_tts_selfhost":
from .sources.gsv_selfhosted_source import (
ProviderGSVTTS as ProviderGSVTTS,
)
case "gsvi_tts_api":
from .sources.gsvi_tts_source import (
ProviderGSVITTS as ProviderGSVITTS,
@@ -0,0 +1,148 @@
import asyncio
import os
import uuid
import aiohttp
from ..provider import TTSProvider
from ..entities import ProviderType
from ..register import register_provider_adapter
from astrbot import logger
from astrbot.core.utils.astrbot_path import get_astrbot_data_path
@register_provider_adapter(
provider_type_name="gsv_tts_selfhost",
desc="GPT-SoVITS TTS(本地加载)",
provider_type=ProviderType.TEXT_TO_SPEECH,
)
class ProviderGSVTTS(TTSProvider):
def __init__(
self,
provider_config: dict,
provider_settings: dict,
) -> None:
super().__init__(provider_config, provider_settings)
self.api_base = provider_config.get("api_base", "http://127.0.0.1:9880").rstrip(
"/"
)
self.gpt_weights_path: str = provider_config.get("gpt_weights_path", "")
self.sovits_weights_path: str = provider_config.get("sovits_weights_path", "")
# TTS 请求的默认参数,移除前缀gsv_
self.default_params: dict = {
key.removeprefix("gsv_"): str(value).lower()
for key, value in provider_config.get("gsv_default_parms", {}).items()
}
self.timeout = provider_config.get("timeout", 60)
self._session: aiohttp.ClientSession | None = None
async def initialize(self):
"""异步初始化:在 ProviderManager 中被调用"""
self._session = aiohttp.ClientSession(
timeout=aiohttp.ClientTimeout(total=self.timeout)
)
try:
await self._set_model_weights()
logger.info("[GSV TTS] 初始化完成")
except Exception as e:
logger.error(f"[GSV TTS] 初始化失败:{e}")
raise
def get_session(self) -> aiohttp.ClientSession:
if not self._session or self._session.closed:
raise RuntimeError(
"[GSV TTS] Provider HTTP session is not ready or closed."
)
return self._session
async def _make_request(
self, endpoint: str, params=None, retries: int = 3
) -> bytes | None:
"""发起请求"""
for attempt in range(retries):
logger.debug(f"[GSV TTS] 请求地址:{endpoint},参数:{params}")
try:
async with self.get_session().get(endpoint, params=params) as response:
if response.status != 200:
error_text = await response.text()
raise Exception(
f"[GSV TTS] Request to {endpoint} failed with status {response.status}: {error_text}"
)
return await response.read()
except Exception as e:
if attempt < retries - 1:
logger.warning(
f"[GSV TTS] 请求 {endpoint}{attempt + 1} 次失败:{e},重试中..."
)
await asyncio.sleep(1)
else:
logger.error(f"[GSV TTS] 请求 {endpoint} 最终失败:{e}")
raise
async def _set_model_weights(self):
"""设置模型路径"""
try:
if self.gpt_weights_path:
await self._make_request(
f"{self.api_base}/set_gpt_weights",
{"weights_path": self.gpt_weights_path},
)
logger.info(f"[GSV TTS] 成功设置 GPT 模型路径:{self.gpt_weights_path}")
else:
logger.info("[GSV TTS] GPT 模型路径未配置,将使用内置 GPT 模型")
if self.sovits_weights_path:
await self._make_request(
f"{self.api_base}/set_sovits_weights",
{"weights_path": self.sovits_weights_path},
)
logger.info(
f"[GSV TTS] 成功设置 SoVITS 模型路径:{self.sovits_weights_path}"
)
else:
logger.info("[GSV TTS] SoVITS 模型路径未配置,将使用内置 SoVITS 模型")
except aiohttp.ClientError as e:
logger.error(f"[GSV TTS] 设置模型路径时发生网络错误:{e}")
except Exception as e:
logger.error(f"[GSV TTS] 设置模型路径时发生未知错误:{e}")
async def get_audio(self, text: str) -> str:
"""实现 TTS 核心方法,根据文本内容自动切换情绪"""
if not text.strip():
raise ValueError("[GSV TTS] TTS 文本不能为空")
endpoint = f"{self.api_base}/tts"
params = self.build_synthesis_params(text)
temp_dir = os.path.join(get_astrbot_data_path(), "temp")
os.makedirs(temp_dir, exist_ok=True)
path = os.path.join(temp_dir, f"gsv_tts_{uuid.uuid4().hex}.wav")
logger.debug(f"[GSV TTS] 正在调用语音合成接口,参数:{params}")
result = await self._make_request(endpoint, params)
if isinstance(result, bytes):
with open(path, "wb") as f:
f.write(result)
return path
else:
raise Exception(f"[GSV TTS] 合成失败,输入文本:{text},错误信息:{result}")
def build_synthesis_params(self, text: str) -> dict:
"""
构建语音合成所需的参数字典。
当前仅包含默认参数 + 文本,未来可在此基础上动态添加如情绪、角色等语义控制字段。
"""
params = self.default_params.copy()
params["text"] = text
# TODO: 在此处添加情绪分析,例如 params["emotion"] = detect_emotion(text)
return params
async def terminate(self):
"""终止释放资源:在 ProviderManager 中被调用"""
if self._session and not self._session.closed:
await self._session.close()
logger.info("[GSV TTS] Session 已关闭")