NoizAI/skills

What it solves

This project provides a central repository of "skills" designed to make AI bots sound more human and natural. It addresses the challenge of creating high-quality, emotionally expressive audio output for AI agents, ranging from simple text-to-speech to complex audio-visual content generation.

How it works

Users can install specific skills using a command-line tool (npx skills add). The repository offers a variety of specialized audio tools, including text-to-speech (TTS) with voice cloning, emotional tuning, and the ability to generate sound effects or full songs with vocals. Some skills can run locally (using the Kokoro backend) or via the Noiz API for enhanced performance and emotion control.

Who it’s for

Developers and AI bot creators who want to add voice capabilities to their agents, as well as users looking to generate high-fidelity audio, music, and dubbed videos for chat platforms like Telegram, Discord, and Feishu.

Highlights

• Local-first security: Run skills on your own machine to keep sensitive data localized.
• Character-style controls: Adjust fillers and emotions to create companion-like speaking styles.
• Multi-platform delivery: Send generated audio directly as native voice messages to Discord, Telegram, and Feishu.
• Diverse audio capabilities: Includes tools for TTS, voice cloning, sound effect generation, text-to-music, and video translation/dubbing.