From 224d59d50d549c3d818631b647e27fade275307e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E6=AC=A3=E5=8D=97=E7=A7=91=E6=8A=80?= Date: Mon, 10 Mar 2025 15:56:28 +0800 Subject: [PATCH] =?UTF-8?q?fix:=E5=A3=B0=E9=9F=B3=E5=8D=A1=E9=A1=BF?= =?UTF-8?q?=E9=97=AE=E9=A2=98=20(#262)?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Co-authored-by: hrz <1710360675@qq.com> --- docs/Deployment.md | 14 +++++------ .../core/handle/sendAudioHandle.py | 25 +++++++++++++------ 2 files changed, 24 insertions(+), 15 deletions(-) diff --git a/docs/Deployment.md b/docs/Deployment.md index 97a11c84..0dfd93b5 100644 --- a/docs/Deployment.md +++ b/docs/Deployment.md @@ -93,7 +93,7 @@ docker logs -f xiaozhi-esp32-server ``` docker stop xiaozhi-esp32-server docker rm xiaozhi-esp32-server -docker rmi ghcr.nju.edu.cn/xinnan-tech/xiaozhi-esp32-server:latest +docker rmi ghcr.nju.edu.cn/xinnan-tech/xiaozhi-esp32-server:server_latest ``` 3、重新按docker方式部署 @@ -299,14 +299,14 @@ LLM: 以下是一些常见问题，供参考： -[1、为什么我说的话，小智识别出来很多韩文、日文、英文](../README.md#1tts-%E7%BB%8F%E5%B8%B8%E5%A4%B1%E8%B4%A5%E7%BB%8F%E5%B8%B8%E8%B6%85%E6%97%B6-) +[1、为什么我说的话，小智识别出来很多韩文、日文、英文](../README.md#1%E4%B8%BA%E4%BB%80%E4%B9%88%E6%88%91%E8%AF%B4%E7%9A%84%E8%AF%9D%E5%B0%8F%E6%99%BA%E8%AF%86%E5%88%AB%E5%87%BA%E6%9D%A5%E5%BE%88%E5%A4%9A%E9%9F%A9%E6%96%87%E6%97%A5%E6%96%87%E8%8B%B1%E6%96%87) -[2、为什么会出现“TTS 任务出错 文件不存在”？](../README.md#1tts-%E7%BB%8F%E5%B8%B8%E5%A4%B1%E8%B4%A5%E7%BB%8F%E5%B8%B8%E8%B6%85%E6%97%B6-) +[2、为什么会出现“TTS 任务出错 文件不存在”？](../README.md#1%E4%B8%BA%E4%BB%80%E4%B9%88%E6%88%91%E8%AF%B4%E7%9A%84%E8%AF%9D%E5%B0%8F%E6%99%BA%E8%AF%86%E5%88%AB%E5%87%BA%E6%9D%A5%E5%BE%88%E5%A4%9A%E9%9F%A9%E6%96%87%E6%97%A5%E6%96%87%E8%8B%B1%E6%96%87) -[3、TTS 经常失败，经常超时](../README.md#1tts-%E7%BB%8F%E5%B8%B8%E5%A4%B1%E8%B4%A5%E7%BB%8F%E5%B8%B8%E8%B6%85%E6%97%B6-) +[3、TTS 经常失败，经常超时](../README.md#1%E4%B8%BA%E4%BB%80%E4%B9%88%E6%88%91%E8%AF%B4%E7%9A%84%E8%AF%9D%E5%B0%8F%E6%99%BA%E8%AF%86%E5%88%AB%E5%87%BA%E6%9D%A5%E5%BE%88%E5%A4%9A%E9%9F%A9%E6%96%87%E6%97%A5%E6%96%87%E8%8B%B1%E6%96%87) -[4、如何提高小智对话响应速度？](../README.md#1tts-%E7%BB%8F%E5%B8%B8%E5%A4%B1%E8%B4%A5%E7%BB%8F%E5%B8%B8%E8%B6%85%E6%97%B6-) +[4、如何提高小智对话响应速度？](../README.md#1%E4%B8%BA%E4%BB%80%E4%B9%88%E6%88%91%E8%AF%B4%E7%9A%84%E8%AF%9D%E5%B0%8F%E6%99%BA%E8%AF%86%E5%88%AB%E5%87%BA%E6%9D%A5%E5%BE%88%E5%A4%9A%E9%9F%A9%E6%96%87%E6%97%A5%E6%96%87%E8%8B%B1%E6%96%87) -[5、我说话很慢，停顿时小智老是抢话](../README.md#1tts-%E7%BB%8F%E5%B8%B8%E5%A4%B1%E8%B4%A5%E7%BB%8F%E5%B8%B8%E8%B6%85%E6%97%B6-) +[5、我说话很慢，停顿时小智老是抢话](../README.md#1%E4%B8%BA%E4%BB%80%E4%B9%88%E6%88%91%E8%AF%B4%E7%9A%84%E8%AF%9D%E5%B0%8F%E6%99%BA%E8%AF%86%E5%88%AB%E5%87%BA%E6%9D%A5%E5%BE%88%E5%A4%9A%E9%9F%A9%E6%96%87%E6%97%A5%E6%96%87%E8%8B%B1%E6%96%87) -[6、我想通过小智控制电灯、空调、远程开关机等操作](../README.md#1tts-%E7%BB%8F%E5%B8%B8%E5%A4%B1%E8%B4%A5%E7%BB%8F%E5%B8%B8%E8%B6%85%E6%97%B6-) +[6、我想通过小智控制电灯、空调、远程开关机等操作](../README.md#1%E4%B8%BA%E4%BB%80%E4%B9%88%E6%88%91%E8%AF%B4%E7%9A%84%E8%AF%9D%E5%B0%8F%E6%99%BA%E8%AF%86%E5%88%AB%E5%87%BA%E6%9D%A5%E5%BE%88%E5%A4%9A%E9%9F%A9%E6%96%87%E6%97%A5%E6%96%87%E8%8B%B1%E6%96%87) diff --git a/main/xiaozhi-server/core/handle/sendAudioHandle.py b/main/xiaozhi-server/core/handle/sendAudioHandle.py index 1cce98b4..8999bf08 100644 --- a/main/xiaozhi-server/core/handle/sendAudioHandle.py +++ b/main/xiaozhi-server/core/handle/sendAudioHandle.py @@ -13,28 +13,37 @@ async def sendAudioMessage(conn, audios, text, text_index=0): logger.bind(tag=TAG).info(f"发送第一段语音: {text}") await send_tts_message(conn, "sentence_start", text) - # 初始化流控参数 - frame_duration = 60 # 毫秒 - start_time = time.perf_counter() # 使用高精度计时器 - play_position = 0 # 已播放的时长（毫秒） + # 流控参数优化 + original_frame_duration = 60 # 原始帧时长（毫秒） + adjusted_frame_duration = int(original_frame_duration * 0.8) # 缩短20% + total_frames = len(audios) # 获取总帧数 + compensation = total_frames * (original_frame_duration - adjusted_frame_duration) / 1000 # 补偿时间（秒） + + start_time = time.perf_counter() + play_position = 0 # 已播放时长（毫秒） for opus_packet in audios: if conn.client_abort: return - # 计算当前包的预期发送时间 + # 计算带加速因子的预期时间 expected_time = start_time + (play_position / 1000) current_time = time.perf_counter() - # 等待直到预期时间 + # 流控等待（使用加速后的帧时长） delay = expected_time - current_time if delay > 0: await asyncio.sleep(delay) - # 发送音频包 await conn.websocket.send(opus_packet) - play_position += frame_duration # 更新播放位置 + play_position += adjusted_frame_duration # 使用调整后的帧时长 + + # 补偿因加速损失的时长 + if compensation > 0: + await asyncio.sleep(compensation) + await send_tts_message(conn, "sentence_end", text) + # 发送结束消息（如果是最后一个文本） if conn.llm_finish_task and text_index == conn.tts_last_text_index: await send_tts_message(conn, 'stop', None)