mirror of
https://github.com/xinnan-tech/xiaozhi-esp32-server.git
synced 2026-07-22 07:03:53 +08:00
Merge branch 'rainv123-patch-1' of https://github.com/xinnan-tech/xiaozhi-esp32-server into rainv123-patch-1
This commit is contained in:
FAN-yeB
+8
-161
@@ -3,14 +3,12 @@
|
||||
## 一、基础环境要求
|
||||
操作系统：Windows / Linux / WSL 2
|
||||
|
||||
Python 版本：3.8 ~ 3.10
|
||||
Python 版本：3.9以上（请根据Paddle官方教程调整）
|
||||
|
||||
Paddle 官方最新版本
|
||||
Paddle 版本：官方最新版本 ```https://www.paddlepaddle.org.cn/install```
|
||||
|
||||
依赖管理工具：conda 或 venv
|
||||
|
||||
前提条件：需要先根据paddle官方教程安装好PaddlePaddle ```https://www.paddlepaddle.org.cn/install```
|
||||
|
||||
## 二、启动paddlespeech服务
|
||||
### 1.从paddlespeech官方仓库拉取源码
|
||||
```bash
|
||||
@@ -18,7 +16,8 @@ git clone https://github.com/PaddlePaddle/PaddleSpeech.git
|
||||
```
|
||||
### 2.建立虚拟环境
|
||||
```bash
|
||||
conda create -n paddle_env python=3.8 -y
|
||||
#请根据Paddle官方支持的python版本建立环境 ```https://www.paddlepaddle.org.cn/install```
|
||||
conda create -n paddle_env python=3.10 -y
|
||||
conda activate paddle_env
|
||||
```
|
||||
### 3.进入paddlespeech目录
|
||||
@@ -46,9 +45,10 @@ paddlespeech tts --input "你好，这是一次测试"
|
||||
### 7.启动服务
|
||||
```yaml
|
||||
paddlespeech_server start --config_file ./demos/streaming_tts_server/conf/tts_online_application.yaml
|
||||
#官方默认启动命令：
|
||||
paddlespeech_server start --config_file ./conf/tts_online_application.yaml
|
||||
```
|
||||
请根据你的```tts_online_application.yaml```的实际目录来启动命令
|
||||
看到如下日志即启动成功
|
||||
请根据你的```tts_online_application.yaml```的实际目录来启动命令，看到如下日志即启动成功
|
||||
```
|
||||
Prefix dict has been built successfully.
|
||||
[2025-08-07 10:03:11,312] [ DEBUG] __init__.py:166 - Prefix dict has been built successfully.
|
||||
@@ -60,160 +60,7 @@ INFO: Uvicorn running on http://0.0.0.0:8092 (Press CTRL+C to quit)
|
||||
|
||||
## 三、修改小智的配置文件
|
||||
### 1.```main/xiaozhi-server/core/providers/tts/paddle_speech.py```
|
||||
```py
|
||||
import asyncio
|
||||
import json
|
||||
import base64
|
||||
import aiohttp
|
||||
import numpy as np
|
||||
import io
|
||||
import wave
|
||||
import websockets
|
||||
from core.providers.tts.base import TTSProviderBase
|
||||
from config.logger import setup_logging
|
||||
|
||||
TAG = __name__
|
||||
logger = setup_logging()
|
||||
|
||||
|
||||
class TTSProvider(TTSProviderBase):
|
||||
def __init__(self, config, delete_audio_file):
|
||||
super().__init__(config, delete_audio_file)
|
||||
self.url = config.get("url", "ws://192.168.1.10:8092/paddlespeech/tts/streaming")
|
||||
self.protocol = config.get("protocol", "websocket")
|
||||
self.spk_id = config.get("spk_id", 0)
|
||||
self.sample_rate = config.get("sample_rate", 24000)
|
||||
self.speed = config.get("speed", 1.0)
|
||||
self.volume = config.get("volume", 1.0)
|
||||
self.save_path = config.get("save_path", "./streaming_tts.wav")
|
||||
|
||||
async def pcm_to_wav(self, pcm_data: bytes, sample_rate: int = 24000, num_channels: int = 1,
|
||||
bits_per_sample: int = 16) -> bytes:
|
||||
"""
|
||||
将 PCM 数据转换为 WAV 文件并返回字节数据
|
||||
:param pcm_data: PCM 数据（原始字节流）
|
||||
:param sample_rate: 音频采样率，默认为24000
|
||||
:param num_channels: 声道数，默认为单声道
|
||||
:param bits_per_sample: 每个样本的位数，默认为16
|
||||
:return: WAV 格式的字节数据
|
||||
"""
|
||||
byte_data = np.frombuffer(pcm_data, dtype=np.int16) # 16位PCM
|
||||
wav_io = io.BytesIO()
|
||||
|
||||
with wave.open(wav_io, "wb") as wav_file:
|
||||
wav_file.setnchannels(num_channels)
|
||||
wav_file.setsampwidth(bits_per_sample // 8)
|
||||
wav_file.setframerate(sample_rate)
|
||||
wav_file.writeframes(byte_data.tobytes())
|
||||
|
||||
return wav_io.getvalue()
|
||||
|
||||
async def text_to_speak(self, text, output_file):
|
||||
if self.protocol == "websocket":
|
||||
return await self.text_streaming(text, output_file)
|
||||
elif self.protocol == "http":
|
||||
return await self.text(text, output_file)
|
||||
else:
|
||||
raise ValueError("Unsupported protocol. Please use 'websocket' or 'http'.")
|
||||
|
||||
async def text(self, text, output_file):
|
||||
request_json = {
|
||||
"text": text,
|
||||
"spk_id": self.spk_id,
|
||||
"speed": self.speed,
|
||||
"volume": self.volume,
|
||||
"sample_rate": self.sample_rate,
|
||||
"save_path": self.save_path
|
||||
}
|
||||
|
||||
try:
|
||||
async with aiohttp.ClientSession() as session:
|
||||
async with session.post(self.url, json=request_json) as resp:
|
||||
if resp.status == 200:
|
||||
resp_json = await resp.json()
|
||||
if resp_json.get("success"):
|
||||
data = resp_json["result"]
|
||||
audio_bytes = base64.b64decode(data["audio"])
|
||||
if output_file:
|
||||
with open(output_file, "wb") as file_to_save:
|
||||
file_to_save.write(audio_bytes)
|
||||
else:
|
||||
return audio_bytes
|
||||
else:
|
||||
raise Exception(
|
||||
f"Error: {resp_json.get('message', 'Unknown error')} while processing text: {text}")
|
||||
else:
|
||||
raise Exception(
|
||||
f"HTTP Error: {resp.status} - {await resp.text()} while processing text: {text}")
|
||||
except Exception as e:
|
||||
raise Exception(f"Error during TTS HTTP request: {e} while processing text: {text}")
|
||||
|
||||
async def text_streaming(self, text, output_file):
|
||||
try:
|
||||
# 使用 websockets 异步连接到 WebSocket 服务器
|
||||
async with websockets.connect(self.url) as ws:
|
||||
# 发送开始请求
|
||||
start_request = {
|
||||
"task": "tts",
|
||||
"signal": "start"
|
||||
}
|
||||
await ws.send(json.dumps(start_request))
|
||||
|
||||
# 接收开始响应并提取 session_id
|
||||
start_response = await ws.recv()
|
||||
start_response = json.loads(start_response) # 解析 JSON 响应
|
||||
if start_response.get("status") != 0:
|
||||
raise Exception(f"连接失败: {start_response.get('signal')}")
|
||||
|
||||
session_id = start_response.get("session")
|
||||
|
||||
# 发送待合成的文本数据
|
||||
data_request = {
|
||||
"text": text,
|
||||
"spk_id": self.spk_id,
|
||||
}
|
||||
await ws.send(json.dumps(data_request))
|
||||
|
||||
audio_chunks = b""
|
||||
timeout_seconds = 60 # 设置超时
|
||||
try:
|
||||
while True:
|
||||
response = await asyncio.wait_for(ws.recv(), timeout=timeout_seconds)
|
||||
response = json.loads(response) # 解析 JSON 响应
|
||||
status = response.get("status")
|
||||
|
||||
if status == 2: # 最后一个数据包
|
||||
break
|
||||
else:
|
||||
# 拼接音频数据（base64 编码的 PCM 数据）
|
||||
audio_chunks += base64.b64decode(response.get("audio"))
|
||||
except asyncio.TimeoutError:
|
||||
raise Exception(f"WebSocket 超时：等待音频数据超过 {timeout_seconds} 秒")
|
||||
|
||||
# 将拼接后的 PCM 数据转换为 WAV 格式
|
||||
wav_data = await self.pcm_to_wav(audio_chunks)
|
||||
|
||||
# 结束请求
|
||||
end_request = {
|
||||
"task": "tts",
|
||||
"signal": "end",
|
||||
"session": session_id # 会话 ID 必须与开始请求中的一致
|
||||
}
|
||||
await ws.send(json.dumps(end_request))
|
||||
|
||||
# 接收结束响应避免服务抛出异常
|
||||
await ws.recv()
|
||||
|
||||
# 返回或保存音频数据
|
||||
if output_file:
|
||||
with open(output_file, "wb") as file_to_save:
|
||||
file_to_save.write(wav_data)
|
||||
else:
|
||||
return wav_data
|
||||
|
||||
except Exception as e:
|
||||
raise Exception(f"Error during TTS WebSocket request: {e} while processing text: {text}")
|
||||
```
|
||||
### 2.```main/xiaozhi-server/data/.config.yaml```
|
||||
使用单模块部署
|
||||
```yaml
|
||||
@@ -237,7 +84,7 @@ python app.py
|
||||
打开test目录下的test_page.html，测试连接和发送消息时paddlespeech端是否有输出日志
|
||||
|
||||
输出日志参考：
|
||||
```b
|
||||
```
|
||||
INFO: 127.0.0.1:44312 - "WebSocket /paddlespeech/tts/streaming" [accepted]
|
||||
INFO: connection open
|
||||
[2025-08-07 11:16:33,355] [ INFO] - sentence: 哈哈，怎么突然找我聊天啦？
|
||||
|
||||
Reference in New Issue
Block a user