claw-voice-local

Convert text to offline Telegram voice messages using piper TTS. Use when the agent should speak a response, send audio, or deliver voice notes via Telegram...

Security Scan

The skill's code matches its stated purpose (local piper TTS → Telegram voice note) but the package metadata and runtime instructions disagree about required credentials/config paths (TELEGRAM_BOT_TOKEN and ~/.openclaw/.env are used though the registry lists none), which is an important incoherence you should resolve before installing.

Purpose & Capability

The scripts (say.py, speak.py, send_voice.py) implement exactly the advertised flow: local piper TTS + ffmpeg → OGG → send to Telegram Bot API. However the registry metadata lists no required environment variables or config paths, while SKILL.md and the code require TELEGRAM_BOT_TOKEN (required), optional TELEGRAM_CHAT_ID, and read a fallback ~/.openclaw/.env. The code also expects a piper binary (~/.local/bin/piper), voice model files (~/.local/share/piper/*.onnx), and ffmpeg — these are all consistent with the stated purpose but the missing declarations in the registry are an inconsistency.

Instruction Scope

Runtime instructions and code stay within scope: they run piper and ffmpeg locally, create temporary WAV/OGG files, and POST the resulting OGG to Telegram's API (api.telegram.org). The skill reads environment variables and a fallback config file (~/.openclaw/.env) for credentials — this is expected for a bot-integrated tool but worth noting since credential files are accessed. No other network endpoints or unexpected data-collection steps are present.

Install Mechanism

There is no automated install spec in the package. README instructs manual downloads from GitHub Releases (piper) and HuggingFace (voice models) and installation of ffmpeg via apt — these are known sources and align with the toolchain. No obscure or shortener URLs, and no opaque archive downloads from personal servers are present.

Credentials

The runtime requires a Telegram bot token (TELEGRAM_BOT_TOKEN) and optionally TELEGRAM_CHAT_ID; these are minimal and appropriate for the task. But the skill registry lists no required environment variables/config paths even though SKILL.md and code use them and read ~/.openclaw/.env as a fallback. That mismatch is material because users might not realize secrets are needed or where they are stored. The credential scope itself is proportional (only the Telegram bot token), but the missing metadata is a red flag.

Persistence & Privilege

The skill does not request 'always: true' or attempt to modify other skills or global agent settings. It runs on demand and uses local files in the user's home directory; it does not create long-lived background services or escalate privileges. It does read/write temporary files and may read ~/.openclaw/.env for credentials, which is normal for a user-space utility.

What to consider before installing

This skill's code appears to implement exactly what it claims: convert text to OGG with local piper + ffmpeg and send a Telegram voice note. However, the skill metadata omitted required credentials and config paths that the code and SKILL.md reference. Before installing or running: - Verify TELEGRAM_BOT_TOKEN is provided and store it securely (use a dedicated bot token and not a highly privileged account). - Ensure the fallback file ~/.openclaw/.env is intentionally used and has restrictive permissions (chmod 600) — review its contents before use. - Confirm you trust the recommended piper and model download sources (GitHub releases and HuggingFace). Download releases directly from the official project pages; verify checksums if available. - Consider creating a dedicated Telegram chat/bot for the skill and limit TELEGRAM_CHAT_ID to the intended recipient. - Because registry metadata omits required env vars, prefer running the scripts manually first (not as an unattended/service skill) and inspect behavior. If you plan to enable it for automatic agent use, add explicit exec allowlist entries and ensure your agent's environment only exposes the minimal required variables. Given the metadata/code mismatch, treat the package as 'suspicious' until the author corrects the declared requirements or you verify them yourself.

Local offline TTS → Telegram Voice Note

Convert any text to a Telegram voice message using piper TTS — fully offline, no cloud API required. Runs on Linux (including Raspberry Pi / ARM64).

When to use

• Send spoken responses instead of text messages
• Read out summaries, alerts, or reports
• Any task where audio feedback is more natural than text

Usage

python3 say.py "Your agent is ready."
python3 say.py "Good morning!" --chat-id 123456789
echo "Task complete." | python3 say.py

Configuration

Set these environment variables (or add to ~/.openclaw/.env):

Variable	Required	Description
TELEGRAM_BOT_TOKEN	Yes	Your Telegram bot token
TELEGRAM_CHAT_ID	Optional	Default chat ID (can be passed as --chat-id)

Environment Variables

TELEGRAM_BOT_TOKEN=required   # Telegram bot token — get one from @BotFather
TELEGRAM_CHAT_ID=optional     # Default target chat ID

Config File

This skill reads ~/.openclaw/.env as a fallback for credentials. Ensure the file has restricted permissions: chmod 600 ~/.openclaw/.env

Installation

See README.md for step-by-step piper installation.

Files

File	Description
say.py	Main entry point: text → Telegram voice note
speak.py	Core TTS: text → OGG Opus via piper + ffmpeg
send_voice.py	Telegram sender: OGG → voice message (no dependencies)

Requirements

• Python 3.11+
• piper binary (see README)
• ffmpeg
• A piper voice model (.onnx)
• Telegram bot token