Telegram Offline Voice

本地生成 Telegram 语音消息，支持自动清洗、分段与临时文件管理。

Security Scan

The skill mostly does what it claims (generate Telegram voice files using ffmpeg and a Python script) but its strong claim of “100% local / zero token” conflicts with its use of the edge-tts library and the install advice that fetches an external installer — this mismatch deserves caution.

Purpose & Capability

Name/description promise: local, offline TTS. Implementation: uses the Python package edge_tts (declared in the script header comments and imported in code). edge_tts typically requests audio from Microsoft TTS endpoints (i.e., network calls), so the claim of being fully offline and zero-token may be false or misleading. Other requirements (ffmpeg, Python, optional uv) match the stated purpose.

Instruction Scope

SKILL.md and the script keep scope to converting provided text -> temporary mp3 -> final ogg and printing file paths. The script only reads the supplied --text and writes temp files under outdir (/tmp by default). No instructions to read unrelated files or environment variables. However, runtime will transitively call edge_tts which may make network requests to Microsoft endpoints — the SKILL.md asserts local-only processing but does not document or justify any network usage.

Install Mechanism

No formal install spec (instruction-only), which is lower risk. SKILL.md recommends installing uv with curl | sh from https://astral.sh/ (third-party installer); piping remote scripts to sh is higher risk and should be treated cautiously. The script lists Python dependencies (edge-tts, aiofiles) in header comments but does not provide explicit pip install steps — user must install these. No arbitrary binary downloads or obscure URLs in included files beyond the uv installer suggestion.

Credentials

The skill requests no environment variables or credentials. The script operates entirely on supplied text and local temp files; there are no demanded tokens, keys, or config path access. This is proportionate to its stated functionality — except for the unresolved offline claim noted above.

Persistence & Privilege

The skill is user-invocable only (always: false) and does not request persistent agent-wide privileges or modify other skills. It does not persist credentials or change system-wide settings.

What to consider before installing

This skill is mostly coherent: it converts text to OGG via edge_tts and ffmpeg and manages temp files correctly. However, the author repeatedly claims "100% local / zero token" while importing and using edge_tts — that library typically obtains TTS audio from Microsoft services (network calls), so the skill may not be truly offline. Before installing or running: 1) Verify edge_tts behavior in your environment (check its docs or run it in a sandbox) to confirm whether audio is generated locally or fetched from the network. 2) If you require strict offline operation, consider replacing edge_tts with a known local TTS engine (Coqui/tts, local OpenTTS endpoint, or local Edge browser invocation) or explicitly audit network traffic. 3) Avoid running curl | sh blindly — the uv install script is optional; prefer installing known packages from distro repos or inspecting the installer first. 4) Run the script in an isolated environment first (container or VM), monitor outbound connections (tcpdump/strace) and review installed Python dependencies (pip show edge-tts) to ensure no unexpected exfiltration. If you cannot confirm edge_tts is purely local, treat the offline claim as inaccurate and proceed with caution.

本地生成，一键封装 — 使用 Microsoft Edge-TTS 生成高质量中文语音，完全离线处理。

💡 为什么需要这个升级？

原生的 TTS 方案通常只能生成 MP3 附件，且无法处理 Markdown 标记和超长文本。本项目通过工程化封装，将“语音合成”进化为“语音交互”：

• 告别标记音：自动识别并清洗 **, #, [link] 等 Markdown 符号，避免 AI 读出这些“代码噪音”。
• 智能对话流：超长文本不再是一读到底，而是按句号、感叹号自动切分为多个语音气泡，听感更像真人在发语音。
• 并发安全：针对多代理/子代理并行调用的场景，使用 UUID 隔离临时文件，彻底杜绝文件读写冲突。
• 零token消耗：完全基于 Edge-TTS 本地生成，无需 OpenAI/Azure 的付费 Token。

✨ 特性

• 🔒 隐私保护：100% 本地音频处理，不经过任何额外云端 TTS 提供商。
• 💰 零token消耗：使用免费的 Edge-TTS 引擎，省下昂贵的 API 额度。
• 🎯 一键生成：内置 voice_gen.py 脚本，自动完成“文本->MP3->OGG”的全过程。
• 🧹 自动清洗：自动剔除 Markdown 符号和 URL 链接，让朗读更自然。
• ✂️ 智能分段：超长文本自动按标点符号切分为多个语音气泡。
• 🛡️ 安全并发：使用 UUID 命名临时文件，支持多代理同时调用。

🛠️ 安装依赖

# 需要 Python 环境和 FFmpeg
sudo apt update && sudo apt install ffmpeg python3-pip -y

# 推荐安装 uv 以极速运行封装脚本
curl -LsSf https://astral.sh/uv/install.sh | sh

🚀 使用方法 (推荐)

直接调用封装脚本，一键生成 Telegram 原生语音气泡路径：

uv run {baseDir}/scripts/voice_gen.py --text "您的待播报内容"

⚙️ 技术细节

参数说明

• --text / -t: 待生成的文本（必填）。
• --voice: 声线选择，默认 zh-CN-XiaoxiaoNeural（晓晓）。
• --rate: 语速调节，默认 +5%。
• --outdir: 临时文件存放目录，默认 /tmp。

自动化清洗规则

脚本会自动移除以下内容以确保朗读流畅：

• Markdown 符号：**, *, _, `, #
• 链接逻辑：[文本](url) 以及所有 http/https 开头的链接
• 分割线：---, *** 等

---

👨💻 关于作者

由 @sanwe 调优并维护。 欢迎关注我的推特获取更多 OpenClaw 进阶玩法：https://x.com/sanwe