agentleFS
jiawood2006jiawood2006/hermes-skills/skills/voice-persona/SKILL.md

voice-persona

Skill2 starsChanged 3 months agoMITView on GitHub

让 Agent 变成能语音对话的机器人:语音文件转文字(支持微信 silk 格式)+ 多音色人格回复(Edge TTS 免费中文音色),全本地零 API 成本。Voice persona chat: transcribe voice messages (incl. WeChat silk) and reply in persona voices.

  • Installs packages

The file

Files
SKILL.md

Voice-Persona · 语音对话 + 人格音色

把 Agent 变成"能听懂语音、会用人格声音回话"的机器人——首个 IM 语音双工闭环技能。

  • 🎙 听懂语音:任意语音文件转文字,支持微信 .silk 格式(独门——大多数开源方案解不了微信语音)
  • 🔊 回语音气泡:人格回复 → Edge TTS 多音色 → silk 编码回发微信语音(pilk 双向,全网首个打通)
  • 🗣 人格回话:内置 6 个人格(元气少女/温柔知心/沉稳大叔/新闻播报/阳光伙伴/随性好友),各配专属中文音色 + 口吻
  • 💰 零 API 成本:faster-whisper(本地)+ Edge TTS(免费)+ pilk(开源)全免费,可选 LLM 增强口吻

Voice chat for agents: transcribe voice (incl. WeChat silk) → reply with persona voices → back to WeChat voice bubbles. First IM two-way voice skill.

与同类差异(2026-09 实测调研)

方案问题
jozhn/wechat-voice-decode-skill只做单向解码转写,无 TTS/人格/语音回复,绑死 openclaw 路径
zhayujie/chatgpt-on-wechat (46k★)语音依赖云端 ASR/TTS,个人微信通道封号风险,无 silk 原生闭环
whisperbot / Telegram 转写 bot 族全部单向"语音→文字",无多音色人格语音回复
SillyTavern (33k★)角色音色强大但只在自家 UI,不接微信/Telegram 语音消息
微软云 voice skills(Azure/OpenAI TTS)单项能力 + 付费云 API,无 IM 场景

别人做"听写"或"朗读",voice-persona 做"在微信/Telegram 里用带人格的嗓音说话"。

快速验证 / Smoke Test(30 秒,免 key)

python3 skills/voice-persona/scripts/voice_persona.py demo
# 🎙 语音 → 文字 → [元气少女/沉稳大叔/新闻播报] 三音色回复音频
# ✅ 全链路通过:语音输入 → 转写 → 人格回复 → 语音输出

安装

# Hermes
hermes skills install jiawood2006/hermes-skills/skills/voice-persona
# 任意环境
git clone https://github.com/jiawood2006/hermes-skills && cd hermes-skills/skills/voice-persona
pip install faster-whisper pilk edge-tts

用法

# 1. 语音 → 文字(微信语音直接传 .silk 文件即可)
python3 voice_persona.py stt wechat_voice.silk
python3 voice_persona.py stt meeting.m4a --model small

# 2. 文字 → 人格回复 → 语音 mp3
python3 voice_persona.py speak "明天记得交报告" --persona yunjian --out reply.mp3

# 3. 只取人格化文本(接你自己的 TTS/IM)
python3 voice_persona.py chat "明天记得交报告" --persona xiaoyi

# 4. 列出人格
python3 voice_persona.py list

# 5. 音频 → 微信 silk 语音(可回发微信语音气泡)
python3 voice_persona.py to_silk reply.mp3 --out reply.silk

# 6. 一条命令双向闭环:人格语音 → 微信格式
python3 voice_persona.py speak "明天早上十点开会" --persona xiaoyi --out r.mp3
python3 voice_persona.py to_silk r.mp3          # → r.silk(#!SILK_V3)

人格库(可扩展)

在 voice_persona.py 顶部 PERSONAS / STYLE_WORDS 增加即可:

key人格Edge TTS 音色口吻
xiaoyi元气少女 · 小伊zh-CN-XiaoyiNeural嘿嘿、啦/呀
xiaoxuan温柔知心 · 晓萱zh-CN-XiaoxuanNeural别担心、慢慢来
yunjian沉稳大叔 · 云健zh-CN-YunjianNeural直说、结论先行
yunyang新闻播报 · 云扬zh-CN-YunyangNeural播报、条理
yunxi阳光伙伴 · 云希zh-CN-YunxiNeural加油、没问题
xiaochen随性好友 · 晓辰zh-CN-XiaochenNeural口语化、像朋友

Agent 接入(Hermes 等框架)

STT 接入:把平台收到的语音消息文件交给 stt 子命令 → 拿到文字进对话流。 Hermes 的 stt.provider: local_command + HERMES_LOCAL_STT_COMMAND 环境变量可直接把微信语音自动接入:

HERMES_LOCAL_STT_COMMAND=<python> <...>/voice_persona.py stt {input_path} --model {model} --output_dir {output_dir} --language {language}

TTS 接入:人格化文本 → speak 出 mp3 → 平台发送语音。

已知陷阱

  • 微信 silk 必须用 pilk 解码:pilk.decode() 输出的是无 RIFF 头的裸 PCM,whisper 读不了——要用 pilk.silk_to_wav()(输出标准 wav,已实测)
  • 首次运行 faster-whisper 会下载模型(small ~500MB),等待 1-3 分钟属正常
  • Edge TTS 需联网(微软免费接口);断网时 speak 会失败,stt/chat 不受影响
  • 转写质量:Intel Mac CPU int8 下 small 模型中文够用;嘈杂/方言建议 base/small 或重发

验证记录(真实产出)

  • 微信 .silk 语音实测转写成功("API这一块我会让海沃云来对接…"39字符/17字符两段)
  • demo 全链路:say 合成中文 → stt 转写 → 3 人格回复 → 3 个 mp3(27-31KB)✅

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.