Hermeneus

From its catalog listing

Live translation for Hermes. Hermeneus captures the audio playing on your Mac, so it works with any app that plays sound. Examples are YouTube, Twitch and X Spaces in the browser, Discord, meeting apps and podcast apps. Verified on real audio: YouTube live in Google Chrome. Captures the audio playing on your Mac (one app, or all system audio except Hermeneus) through a Core Audio process tap on macOS 14.2+, transcribes it with ElevenLabs Scribe v3 Live, translates each line with your own Hermes model and shows live subtitles in a Desktop pane, a floating overlay above other apps, the terminal (hermes hermeneus run), /translate, or an opt-in 127.0.0.1 page for OBS; spoken translation with ElevenLabs Flash v2.5 is opt-in, and language, mode, voice, output and source change while a session runs. Listening covers 100 languages and speaking covers 32.

Good to know before installing

streams the captured audio to ElevenLabs (wss://api.elevenlabs.io) with your ELEVENLABS_API_KEY while a session runs (it stops sending after about 10 s of digital silence and resumes with the first sound), and sends the translated text there too when spoken translation is on; lists your account's voices with GET https://api.elevenlabs.io/v2/voices (voices command and the pane's voice picker); sends each committed line plus up to 3 earlier lines (context_segments, 0 to 10) to your configured model through ctx.llm under its own auxiliary task, live_translate (doctor --test-model sends one tiny extra request); /translate start and /translate set --source are refused from chat platforms unless allow_remote_start is on (default off); needs the macOS System Audio Recording permission for the app that runs Hermes (Microphone for the fallback); the setup's Allow system audio and Check again, the doctor, and the pane's Doctor tab every time it opens, play a 440 Hz test tone for about half a second through a short-lived player subprocess, on BlackHole or VB-Cable when installed, otherwise quietly on the default output while the tap mutes it (--no-audio-check skips it); the process tap runs one capture worker subprocess per session with a minimal environment, stopped with the session, on unload and at exit; the floating overlay is a separate Python process while it is open, with the same minimal environment, and keeps six files under plugin-data/hermeneus/overlay/: lock, settings.json, theme.json, positions.json, positions.json.lock (empty, locked while positions.json is written) and overlay.log; positions.json is written through a short-lived .positions-*.tmp file in the same folder, by the overlay and by the process that serves the pane's overlay routes; the pane's first-run setup writes plugin-data/hermeneus/setup.json (the last audio check result, no audio, no key) and audio-check.running while that check captures (removed when it ends); settings.json, theme.json and setup.json are each written to a short-lived .tmp file beside them and renamed over the file at once; settings you change (in the pane, in Settings or with hermes hermeneus overlay-page) are saved by Hermes in its config.yaml; one session at a time on its own thread, reached through a Unix domain socket (mode 0600) under its plugin-data; no TCP port unless you turn on the overlay page, which listens on 127.0.0.1:47821 (overlay_page_port) only, behind a per-session token and a Host check; runs /usr/sbin/system_profiler SPAudioDataType and /bin/ps (read only, absolute paths, argument lists, minimal environment) to check audio routes and name the app in the permission message, and /usr/bin/open with the System Settings privacy address (x-apple.systempreferences:...Privacy_ScreenCapture) when you press Open System Settings in the setup, which changes nothing; the fallback reads a virtual device (BlackHole 2ch or VB-Cable) that you install yourself, never bundled or downloaded; subtitles stay in memory (last 200 lines) and no audio or transcript is written to disk; no telemetry. Platform status: macOS is tested on real audio; Windows support is implemented and reviewed against documentation, not yet run on a Windows machine, so this entry lists macOS only (Windows preview: docs/windows.md).

From its catalog listing.

How to install

  1. Install it. Hermes asks for any API keys it needs (this one needs ELEVENLABS_API_KEY), then asks whether to turn it on.
    hermes plugins install hermeneus
  2. Said no? Turn it on later:
    hermes plugins enable hermeneus
  3. Restart Hermes so it loads the plugin.

Some plugins also need a sign-in or settings: see Full setup & settings. Memory plugins ask whether to become your memory provider instead. Steps follow the official plugins guide (opens in a new tab).

Category
Chat apps & voice
Maintained by
Bihruze (Community)
Needs
ELEVENLABS_API_KEY API key
Works on
macos
Needs Hermes
0.23.1 or newer
Version
0.1.0
In the catalog
2026-11-11
GitHub stars
2
Using the Hermes desktop app? See where plugins live
Hermes desktop Capabilities page, Plugins tab filtered to Official plugins built by Nous Research, each with an add button.
The official plugin catalog built into the app.Hermes desktop v0.22.5 · Capabilities > Plugins · Windows · captured Nov 1, 2026

Full setup & settings ↗Source code ↗

More in chat apps & voice

All chat apps & voice plugins

New Hermes Agent updates every day on X. Follow @HermesWatcher

← All plugins