voicedbyme Download

VoicedByMe / Use cases / Hermes Agent voice

Hermes Agent · Local TTS · Voice mode · MCP

Hermes Agent, in your voice.

VoicedByMe gives Hermes Agent a voice that runs on your own computer. One click adds hooks so Hermes says out loud when it needs your approval or has a question, plus a speak tool. And because VoicedByMe serves an OpenAI-compatible speech endpoint, Hermes's voice mode can answer in your cloned voice, offline, instead of a cloud voice.

Download for Windows Download for Mac Free for a day · then $30 once

Samples made offline with ready-made voices. All of them are AI-generated.

How it works

Connect once. Restart Hermes.

  1. Install VoicedByMe

    Windows 10 and 11 or a Mac with Apple silicon. Pick a ready-made voice or clone your own from twenty seconds of talking.

  2. Connect Hermes

    Open AI agents and press Connect. It finds Hermes's config.yaml, backs it up, and adds a speak tool and two hooks, leaving your comments and layout alone.

  3. Restart and approve

    Restart Hermes and approve VoicedByMe's hooks when it asks. From then on, Hermes speaks up when it's waiting on you.

What it says

Hermes, when it needs you.

Approval requests

"Hermes wants to run rm -rf build. Allow it?" The command it wants approved, read before you even look at the screen.

Questions

"Hermes has a question. Which branch should I deploy?" Its question, word for word.

Speaking on request

"Say good morning in my voice" works: Hermes gets a speak tool over MCP and knows when to use it.

A voice per agent

Hermes can sound different from Claude Code and Codex, so you know who's asking.

Nothing changes

The hooks only read what Hermes is waiting for and never answer for you. Disconnect removes exactly the lines it added.

Quiet hours

No alerts at night, and quiet while the terminal Hermes runs in is in front of you.

Local TTS

Your own voice as Hermes's TTS provider.

Hermes reads its replies aloud in voice mode with whichever TTS provider you set. Its openai provider accepts a base_url for OpenAI-compatible servers, and VoicedByMe is one, running on your computer. Point Hermes at it and every spoken reply is in your cloned voice, or any of the 70 ready-made voices, without a cloud voice or a per-character bill.

1. Copy the address and key

In VoicedByMe, open AI agents, then Other apps and scripts. Copy the address (it looks like http://127.0.0.1:47821/v1) and the key.

2. Set the provider in config.yaml

tts:
  provider: "openai"
  openai:
    base_url: "http://127.0.0.1:47821/v1"
    api_key: "vr-…"        # the key VoicedByMe shows
    voice: "Maya"          # any VoicedByMe voice, or your own voice's name

3. Turn on voice mode

In Hermes, run /voice on, then /voice tts to hear every reply. Hermes streams speech a sentence at a time as 24 kHz audio, which is what VoicedByMe makes, so the first words come quickly.

Two limits, to be straight: voice notes for Telegram, WhatsApp and Signal ask for Opus audio, which VoicedByMe doesn't make, so keep a different provider for those. And speech recognition is separate: Hermes's local Whisper keeps that side offline too.

Compared

Ways to give Hermes a voice.

OptionRuns offlineYour own voiceSays approval requests and questionsSetup
VoicedByMeYesYes, from 20 secondsYesOne click, plus three lines for voice mode
Edge TTS (Hermes's default)No, uses Microsoft's online voicesNoNoBuilt in
ElevenLabs or OpenAI TTSNoWith ElevenLabs voice cloningNoAn API key, paid per use
Piper, KittenTTS, NeuTTS, KokoroYesNeuTTS, from a reference clipNoPython packages or a local server

As of October 2026, Hermes Agent v0.21. Hermes's voice mode uses whichever provider you set; VoicedByMe's hooks work with any of them.

Questions

How do I give Hermes Agent a voice?

Two ways that work together. Connect VoicedByMe in its AI agents page so Hermes says when it needs approval or has a question, and set VoicedByMe as Hermes's openai TTS provider so voice mode answers in a natural voice or yours.

Can Hermes Agent use a local TTS?

Yes. Hermes has local providers built in (Piper, KittenTTS and NeuTTS), and its openai provider can point at any OpenAI-compatible server through tts.openai.base_url: VoicedByMe, or a Kokoro server. VoicedByMe adds your own cloned voice and a normal installer.

Can Hermes speak in my own voice?

Yes. Record about twenty seconds in VoicedByMe, then put that voice's name in tts.openai.voice. Hermes's spoken replies, approval requests and questions are then in your voice.

Does Hermes voice mode work offline?

Speech recognition can run locally with Whisper, and with VoicedByMe the speech does too. The language model is the one part that needs the internet, unless you run that locally as well.

Why does Hermes ask me to approve the hooks?

Hermes asks before it runs new shell hooks, which is a good safety check. Approve VoicedByMe's once after restarting; they only read what Hermes is waiting for and say it.

Is anything sent online?

Not by VoicedByMe. The voice runs on your computer and only takes requests from programs on it, with a private key.

Which other agents can speak?

Claude Code, Codex, Cursor and the Claude app connect with one click too. See Claude Code voice and Codex voice.

How much does it cost?

$30 once for the first 50 customers, then $50. No subscription. The full app is free for one day so you can try it first. A license is for one computer.

For Windows 10 and 11, and Macs with Apple silicon

Hermes, in your voice.

$30once. Launch price for the first 50 customers, then $50.
One computer. No subscription.

Download for Windows Download for Mac Also for Windows 10 and 11 Also for Macs with Apple silicon

The full app, free for 24 hours. Purchases aren't refundable, so please try it first.