mirror of
https://github.com/Ethan930717/siliconflow_tts.git
synced 2026-07-21 22:53:59 +08:00
246 lines
9.5 KiB
Python
246 lines
9.5 KiB
Python
"""硅基流动TTS组件"""
|
||
import asyncio
|
||
import logging
|
||
import aiohttp
|
||
import async_timeout
|
||
import base64
|
||
import os
|
||
from typing import Any
|
||
|
||
from homeassistant.components.tts import (
|
||
TextToSpeechEntity,
|
||
Voice,
|
||
TtsAudioType,
|
||
)
|
||
from homeassistant.config_entries import ConfigEntry
|
||
from homeassistant.core import HomeAssistant
|
||
from homeassistant.helpers.entity_platform import AddEntitiesCallback
|
||
|
||
from .const import (
|
||
API_ENDPOINT,
|
||
DEFAULT_MODEL,
|
||
DEFAULT_VOICE,
|
||
DEFAULT_RESPONSE_FORMAT,
|
||
DEFAULT_SAMPLE_RATE,
|
||
DEFAULT_SPEED,
|
||
DEFAULT_GAIN,
|
||
DEFAULT_STREAM,
|
||
SUPPORTED_FORMATS,
|
||
MIN_SPEED,
|
||
MAX_SPEED,
|
||
MIN_GAIN,
|
||
MAX_GAIN,
|
||
get_voice_options,
|
||
CONF_VOICE,
|
||
CONF_SPEED,
|
||
CONF_SAMPLE_RATE,
|
||
CONF_RESPONSE_FORMAT,
|
||
CONF_GAIN,
|
||
)
|
||
|
||
_LOGGER = logging.getLogger(__name__)
|
||
|
||
# 支持的语音列表（根据CosyVoice2-0.5B模型的实际语音）
|
||
SUPPORTED_VOICES = {
|
||
"zh": [
|
||
Voice("speech:Dahu-HA-Lulu:qc10fwc040:ugmvnfqyesexvqjuneyf", "Lulu"),
|
||
],
|
||
"en": [
|
||
Voice("speech:Dahu-HA-Lulu:qc10fwc040:ugmvnfqyesexvqjuneyf", "Lulu"),
|
||
]
|
||
}
|
||
|
||
async def async_setup_entry(
|
||
hass: HomeAssistant,
|
||
config_entry: ConfigEntry,
|
||
async_add_entities: AddEntitiesCallback,
|
||
) -> None:
|
||
"""设置TTS实体"""
|
||
async_add_entities([SiliconFlowTTSEntity(config_entry)])
|
||
|
||
class SiliconFlowTTSEntity(TextToSpeechEntity):
|
||
"""硅基流动TTS实体"""
|
||
|
||
def __init__(self, config_entry: ConfigEntry) -> None:
|
||
"""初始化TTS实体"""
|
||
self._config_entry = config_entry
|
||
self._attr_name = "SiliconFlow TTS"
|
||
self._attr_unique_id = f"siliconflow_tts_{config_entry.entry_id}"
|
||
|
||
@property
|
||
def supported_languages(self) -> list[str]:
|
||
"""返回支持的语言列表"""
|
||
return list(SUPPORTED_VOICES.keys())
|
||
|
||
def get_supported_voices(self, language: str) -> list[Voice] | None:
|
||
"""获取指定语言支持的语音（动态加载所有可用音色）"""
|
||
# 从动态音色选项中生成Voice对象列表
|
||
voice_options = get_voice_options()
|
||
voices = []
|
||
for voice_id, voice_name in voice_options.items():
|
||
voices.append(Voice(voice_id, voice_name))
|
||
_LOGGER.debug(f"为语言 '{language}' 提供 {len(voices)} 个音色选项")
|
||
return voices
|
||
|
||
@property
|
||
def default_language(self) -> str:
|
||
"""返回默认语言"""
|
||
return "zh"
|
||
|
||
@property
|
||
def supported_audio_codecs(self) -> list[str]:
|
||
"""返回支持的音频编解码器"""
|
||
return SUPPORTED_FORMATS
|
||
|
||
def _encode_audio_file(self, file_path: str) -> str | None:
|
||
"""将音频文件编码为base64字符串"""
|
||
try:
|
||
if not os.path.exists(file_path):
|
||
_LOGGER.error(f"音频文件不存在: {file_path}")
|
||
return None
|
||
|
||
with open(file_path, 'rb') as audio_file:
|
||
audio_data = audio_file.read()
|
||
|
||
# 检查文件大小，如果太大则截取前100KB
|
||
if len(audio_data) > 100 * 1024: # 100KB
|
||
_LOGGER.warning(f"音频文件较大 ({len(audio_data)} bytes)，截取前100KB")
|
||
audio_data = audio_data[:100 * 1024]
|
||
|
||
encoded_audio = base64.b64encode(audio_data).decode('utf-8')
|
||
_LOGGER.debug(f"成功编码音频文件: {file_path}, 大小: {len(audio_data)} bytes")
|
||
return encoded_audio
|
||
except Exception as e:
|
||
_LOGGER.error(f"编码音频文件失败: {e}")
|
||
return None
|
||
|
||
def _is_uri_reference(self, reference: str) -> bool:
|
||
"""检查是否为URI格式的参考音频"""
|
||
return reference.startswith("speech:")
|
||
|
||
async def async_get_tts_audio(
|
||
self,
|
||
message: str,
|
||
language: str,
|
||
options: dict[str, Any] | None = None,
|
||
) -> TtsAudioType:
|
||
"""获取TTS音频"""
|
||
|
||
api_key = self._config_entry.data.get("api_key")
|
||
if not api_key:
|
||
_LOGGER.error("未找到API密钥")
|
||
return None
|
||
|
||
try:
|
||
headers = {
|
||
"Authorization": f"Bearer {api_key}",
|
||
"Content-Type": "application/json",
|
||
}
|
||
|
||
# 从配置条目的选项中获取参数，优先级：service call options > config entry options > defaults
|
||
config_options = self._config_entry.options
|
||
_LOGGER.debug(f"配置选项: {config_options}")
|
||
_LOGGER.debug(f"服务调用选项: {options}")
|
||
|
||
# 优化参数获取逻辑，确保正确处理None值
|
||
response_format = (
|
||
options.get("response_format") if options and "response_format" in options
|
||
else config_options.get(CONF_RESPONSE_FORMAT, DEFAULT_RESPONSE_FORMAT)
|
||
)
|
||
speed = (
|
||
options.get("speed") if options and "speed" in options
|
||
else config_options.get(CONF_SPEED, DEFAULT_SPEED)
|
||
)
|
||
sample_rate = (
|
||
options.get("sample_rate") if options and "sample_rate" in options
|
||
else config_options.get(CONF_SAMPLE_RATE, DEFAULT_SAMPLE_RATE)
|
||
)
|
||
gain = (
|
||
options.get("gain") if options and "gain" in options
|
||
else config_options.get(CONF_GAIN, DEFAULT_GAIN)
|
||
)
|
||
|
||
# 验证参数范围
|
||
speed = max(MIN_SPEED, min(MAX_SPEED, speed))
|
||
gain = max(MIN_GAIN, min(MAX_GAIN, gain))
|
||
|
||
# 构建输入文本，根据API文档添加特殊标记，并解决语音切断问题
|
||
input_text = message.strip()
|
||
|
||
# 解决语音切断问题：在文本末尾添加适当的标点符号和停顿
|
||
if not input_text.endswith(('.', '。', '!', '！', '?', '？', '，', ',', '、')):
|
||
input_text += "。" # 添加句号确保语音完整
|
||
|
||
# 添加轻微的停顿来避免切断（使用标点符号）
|
||
input_text += " " # 两个空格提供自然的停顿
|
||
|
||
if options and "emotion" in options:
|
||
emotion = options["emotion"]
|
||
input_text = f"Can you say it with a {emotion} emotion? <|endofprompt|>{input_text}"
|
||
|
||
_LOGGER.debug(f"处理后的文本: '{input_text}'")
|
||
|
||
# 构建payload
|
||
payload = {
|
||
"model": DEFAULT_MODEL,
|
||
"input": input_text,
|
||
"response_format": response_format,
|
||
"speed": speed,
|
||
"sample_rate": sample_rate,
|
||
"gain": gain,
|
||
"stream": DEFAULT_STREAM, # 显式设置为false，确保最快响应
|
||
}
|
||
|
||
# 确定使用的语音，优先级：service call options > config entry options > defaults
|
||
voice = DEFAULT_VOICE
|
||
|
||
# 检查是否在service call中指定了语音（包括预设语音和自定义音色）
|
||
if options and "voice" in options:
|
||
voice = options["voice"]
|
||
_LOGGER.info(f"使用service call指定的语音: {voice}")
|
||
elif options and "reference_audio" in options:
|
||
# 为了兼容性，仍然支持reference_audio参数
|
||
reference_audio = options["reference_audio"]
|
||
if self._is_uri_reference(reference_audio):
|
||
voice = reference_audio
|
||
_LOGGER.info(f"使用reference_audio参数指定的URI音色: {reference_audio}")
|
||
else:
|
||
_LOGGER.warning(f"本地文件参考音频暂不支持，请先上传获取URI: {reference_audio}")
|
||
elif config_options.get(CONF_VOICE):
|
||
# 使用配置的音色
|
||
voice = config_options.get(CONF_VOICE)
|
||
_LOGGER.info(f"使用配置的语音: {voice}")
|
||
else:
|
||
# 使用全局默认语音
|
||
_LOGGER.info(f"🎭 使用全局默认语音: {voice}")
|
||
|
||
# 验证选择的音色是否在可用音色列表中
|
||
voice_options = get_voice_options()
|
||
if voice not in voice_options:
|
||
_LOGGER.warning(f"选择的音色 '{voice}' 不在可用列表中，使用默认音色")
|
||
voice = DEFAULT_VOICE
|
||
|
||
payload["voice"] = voice
|
||
|
||
_LOGGER.info(f"TTS配置 - 音色:{voice[-10:]}..., 格式:{response_format}, 采样率:{sample_rate}Hz, 语速:{speed}, 增益:{gain}dB")
|
||
_LOGGER.debug(f"完整音色ID: {voice}")
|
||
_LOGGER.debug(f"完整payload: {payload}")
|
||
|
||
async with aiohttp.ClientSession() as session:
|
||
async with async_timeout.timeout(30):
|
||
async with session.post(API_ENDPOINT, headers=headers, json=payload) as response:
|
||
if response.status == 200:
|
||
audio_data = await response.read()
|
||
_LOGGER.debug(f"TTS成功，音频大小: {len(audio_data)} bytes")
|
||
return (response_format, audio_data)
|
||
else:
|
||
error_text = await response.text()
|
||
_LOGGER.error(f"TTS API错误 {response.status}: {error_text}")
|
||
return None
|
||
|
||
except asyncio.TimeoutError:
|
||
_LOGGER.error("TTS请求超时")
|
||
return None
|
||
except Exception as e:
|
||
_LOGGER.error(f"TTS处理错误: {e}")
|
||
return None |