Hermes Voicebox: plugin for Hermes Agent

Voicebox text-to-speech and Whisper speech-to-text providers for Hermes Agent, fully local

What Hermes Voicebox does

Hermes Voicebox adds Voicebox as a text-to-speech and speech-to-text provider for Hermes Agent. Hermes already speaks replies and transcribes voice memos, and it ships with cloud defaults: Microsoft Edge TTS for speaking and Groq or OpenAI Whisper for transcribing. This plugin adds no new features; it reroutes that existing pipeline through the Voicebox app on your machine, so spoken replies in the CLI, TUI and chat voice bubbles use a cloned or preset Voicebox voice and incoming audio is transcribed locally.

Because transcription happens before the model runs and synthesis after it finishes, this is pipeline plumbing rather than a tool the model calls. Voicebox's separate MCP server covers deliberate actions, such as reading text in a chosen voice or recalling dictated captures, and the README suggests connecting both. A bundled skill teaches the agent when speaking aloud is appropriate and how to pick voices. You select the provider with tts.provider and stt.provider set to voicebox in config.yaml.

Key features

Voicebox as the Hermes TTS provider, with cloned or preset voices
Local Whisper transcription for voice messages and push-to-talk recordings
Works for CLI and TUI speech and voice bubbles on Telegram, Discord, WhatsApp, Slack and Signal
Bundled skill for choosing voices and knowing when to speak aloud
Voice selection by Voicebox profile name or id, with engine ids such as qwen, kokoro and chatterbox
Optional pairing with Voicebox's MCP server for agent-invoked voice actions

When to use it

Replying to Telegram voice memos without sending audio to Groq or OpenAI
Giving your Hermes agent a cloned voice for spoken replies
Keeping the whole Hermes voice pipeline on your own machine

Who it is for: Hermes Agent users who want voice features to run locally with Voicebox instead of cloud speech services.

How it fits with Hermes Agent

Built for Hermes Agent as a plugin that Hermes finds through the hermes_agent.plugins entry point, replacing the engine behind its existing TTS and STT settings.

How to install Hermes Voicebox

These commands are copied from the project's README. Check the repository for the latest steps before you run them.

pip install hermes-voicebox

Requirements: Voicebox running locally (the desktop app, or Docker on 127.0.0.1:17600), Hermes Agent and Python 3.11+

FAQ

What is Hermes Voicebox?

Hermes Voicebox is a plugin that provides Voicebox text-to-speech and speech-to-text engines to Hermes Agent. It enables voice cloning and local Whisper transcription.

Does Hermes Voicebox add new tools to Hermes Agent?

No. The README says it adds no new features and only swaps the engine behind existing voice features. For agent-chosen voice actions, connect Voicebox's separate MCP server as well.

How do I install Hermes Voicebox?

Run pip install hermes-voicebox in the environment Hermes runs in, then set tts.provider and stt.provider to voicebox in ~/.hermes/config.yaml. A directory install with hermes plugins enable voicebox is also described.