TTS Tool Skill

tts-tool is an installable Agent Skill for turning text or markdown into pleasant speech audio with chunktts.

Features

It provides one workflow:

• rewrites symbol-heavy or markdown-heavy text into natural spoken text
• inserts <bk> markers at speech-friendly boundaries
• runs chunktts
• returns one final .opus file and its location

This skill is designed for content that would otherwise sound bad in TTS, such as raw URLs, markdown syntax, LaTeX, file paths, shell commands, and technical notation.

Requirements

• chunktts: Required to generate the final .opus file.
• Runtime libraries: Linux requires libcurl and libsndfile; macOS requires curl and libsndfile. Windows release archives bundle the required DLLs.
• DeepInfra API Key: Required by chunktts.
  • Set it via DEEPINFRA_API_KEY (recommended).
  • Or provide it via config.json next to the chunktts executable.

Installation

Using Codex

Codex recommends installing non-built-in skills using the $skill-installer. Prompt Codex with:

$skill-installer install the skill from repo planetis-m/tts-tool with path .

Manual Install

Clone directly into your agent's scanned skills path:

git clone https://github.com/planetis-m/tts-tool.git ~/.agents/skills/tts-tool

Usage Examples

Invoke the skill explicitly using $tts-tool in your prompts:

Use $tts-tool to turn this markdown into a pleasant spoken .opus file.

Use $tts-tool on notes.md and generate the final audio.