Hermes Agent ElevenLabs support now goes way past voices: there's an official ElevenLabs Sound Effects plugin that turns a plain text description into a sound file, right inside your chat. Install it with one command, add your ElevenLabs API key, and ask for "a door creaking open in a stone hall". Hermes makes it, saves it and sends it to you.
📺 Watch: NEW Hermes Voice Agent Is CRAZY!
🔥 Get the Agent OS as a free bonus: AI Profit Boardroom members get the full Agent OS zip, prompt libraries, daily tutorials and weekly live coaching calls. → Get inside
Nous Research announced it on X on 27 October 2026: generate SFX with ElevenLabs straight from Hermes Desktop, with looped playback, one-click variations and a library of all your sounds. This guide covers the plugin, the install, the settings, and the other ElevenLabs routes Hermes already had for speaking and listening.
What the ElevenLabs Sound Effects plugin does
The plugin is called elevenlabs-sound-effects. On the Hermes plugin catalog it's marked Official, sits under Tools, is published by NousResearch, and was added on 26 October 2026 at version 1.2.0.
It gives the agent one tool: elevenlabs_sound_effect. You describe a sound, and the tool turns it into an audio file between 0.5 and 30 seconds long. It can also make a seamless loop, which is the bit game builders and video editors care about.
Per the README, three more pieces ship with it:
- A sound-design skill. It teaches the agent how to write ElevenLabs prompts (based on ElevenLabs' own prompting guide), make loops, pick a format for where the sound is going, and build a sound pack.
- A local ffmpeg script. sfx_tool.py does mono downmix, trims silence, normalises loudness, adds fades, checks loop seams for clicks, converts formats and packs a set into a folder with a manifest and a zip. The README says this costs no ElevenLabs credits.
- Desktop extras. An inline sound card in the chat and a sound library tab.
Teknium framed it on X as part of a wider suite of Game Dev Studio capabilities for Hermes. If you're building game assets, pair it with the 3D side in my Hermes Meshy guide and you've got models and sound from one agent.
I run Hermes every day inside my Agent OS with a profile per model, and plugins like this are why. One agent, one chat, and it keeps picking up new skills. Want the full setup? The AI Profit Boardroom ($69/mo) gives you the Agent OS, a 30-day AI agent roadmap, daily tutorials as new tools drop, 4 coaching calls a week and the prompt library. 3,400+ members are in there.
📺 Watch: New Hermes Elevenlabs Voice Agent Is INSANE!
How to install Hermes Agent ElevenLabs sound effects
Three steps, all from the plugin README.
- Install the plugin. Run hermes plugins install elevenlabs-sound-effects. You can also install straight from the repo with hermes plugins install NousResearch/hermes-official-plugins#elevenlabs-sound-effects.
- Add your API key. The installer asks for your ElevenLabs key if ELEVENLABS_API_KEY isn't set yet. Create one under Profile, then API keys, in your ElevenLabs account settings.
- Ask for a sound. The tool only shows up for the agent once the key is set. Then just describe what you want.
Already using ElevenLabs for Hermes voice? It's the same key, so there's nothing new to enter. One catch: if you use a restricted key, switch on the Sound Effects permission. It's separate from Text to Speech.
Plugins install per profile. If you run several profiles like I do, add it to another one with hermes -p <profile> plugins install elevenlabs-sound-effects.
Not sure where sound fits in your workflow? Book a free AI strategy session and we'll map it out.
📺 Watch: Hermes3D Is INSANE: AI Agents Now Have a 3D Office
Turning on the Desktop sound card and library
This is the part from the Nous announcement, and it's opt-in. Installing from the CLI only turns on the agent half.
To switch on the Desktop half, open Capabilities, then Plugins, go to the list view, and turn on "Desktop: ElevenLabs Sound Effects" on the plugin's row. Without it, Desktop still plays every sound in its built-in player.
With it on, each sound you generate shows up as a card in the chat:
- Waveform and playback. Click to seek, play and pause.
- Loop toggle. Repeats the sound gaplessly so you can hear the seam.
- Download. The original, or MP3, WAV or OGG converted by ffmpeg on the machine Hermes runs on.
- Next. Three variations, shorter, longer, make it loop, or try the other model. Each row shows its cost based on what ElevenLabs billed, and nothing is billed until the agent actually runs the tool.
The library opens from the command palette (Cmd+K, then "ElevenLabs: open sound library"). It lists every sound the active profile made, with its prompt and settings, a format filter and a prompt search. You can reuse a prompt or jump back to the chat that made it.
Two version notes from the README. The sound card needs a Hermes Desktop with tool-media renderers, first shipping in the release after v0.22.5. On older Desktops the same sound just plays in the built-in player. Running Hermes on a remote machine? Install the plugin there, then in your Desktop use Capabilities, Plugins, Install from Git with NousResearch/hermes-official-plugins/elevenlabs-sound-effects and Desktop UI checked.
Settings worth changing
In Desktop, open Capabilities, Plugins, then the gear icon on the plugin. Everywhere else, use hermes config set. The main settings:
| Setting | Options | What it changes |
|---|---|---|
| sound_model | eleven_text_to_sound_v2 or eleven_text_to_sound_v3 | Which ElevenLabs sound model runs by default |
| output_format | mp3_44100_128, mp3_44100_192, opus_48000_128, opus_48000_192, wav_44100, wav_48000 | File type and quality |
| api_base_url | https://api.elevenlabs.io by default | Only change it for an ElevenLabs data-residency endpoint |
| api_key | Stored in .env as ELEVENLABS_API_KEY | Your ElevenLabs account |
Example: hermes config set plugins.entries.elevenlabs-sound-effects.settings.sound_model eleven_text_to_sound_v3. Heads up on formats: per the README, ElevenLabs lists 192 kbps MP3 for its Creator plan and above, and 44.1 kHz WAV for Pro and above. If your plan refuses a format, the tool passes the error back.
The agent picks the per-sound details itself: the description (450 characters max), the length (0.5 to 30 seconds, or leave it to ElevenLabs), loop on or off, and prompt influence from 0 to 1. Higher sticks closer to your words. Lower gives more variety.
Where do the files go? Each sound saves to the generated audio folder inside your Hermes home and never overwrites an older one. The plugin also logs each generation, including the credits billed, to its own data folder. The README says nothing in that log leaves your machine.
Other ElevenLabs routes in Hermes: voice, TTS and transcription
People also search "hermes voice elevenlabs", and that's a separate feature that was there before this plugin. The Hermes docs list ElevenLabs as a built-in provider for both speaking and listening.
ElevenLabs as the TTS voice
Set tts.provider to elevenlabs in config.yaml and put ELEVENLABS_API_KEY in your .env file. The default voice is Adam with the eleven_multilingual_v2 model. The docs mention eleven_v3 and eleven_flash_v2_5 as options, and Desktop's Settings, Voice screen accepts any model id.
The neat bit: with ElevenLabs as the provider, the agent can pick a different voice per reply. Its text_to_speech tool gains a voice_id field and a find_voices search across the public Voice Library or your own account, filtered by gender, age, accent and language. Each result has a preview link so you can hear it first. The docs note library voices only work over the API on paid ElevenLabs plans, and the key needs Text to Speech plus Voices: Read permissions.
ElevenLabs for speech-to-text
ElevenLabs Scribe is one of the built-in transcription providers, so voice notes from Telegram, Discord and the rest can be transcribed through it. Live transcription in voice mode uses Scribe v2 realtime. For streaming speech, the auto setting tries ElevenLabs first, then Gemini, OpenAI and xAI.
Want more plugins like this one? I keep a running list of the best Hermes Agent plugins. For picking the model that drives your agent, check Goldie Bench, and the Agent OS guide shows how I organise profiles. Prefer to plan it with me? Grab a free AI strategy session.
FAQ
Do I need a paid ElevenLabs plan for sound effects?
You need an ElevenLabs API key. Some output formats are plan-gated per the README: 192 kbps MP3 from Creator up, 44.1 kHz WAV from Pro up. Check your plan on ElevenLabs before picking those.
Does the sound effects plugin work outside Hermes Desktop?
Yes. The sound is sent as an attachment on Telegram, Discord, Slack and other messaging platforms, and in the terminal you get the file path. The card and library are the Desktop-only extras.
Is this the same as the Hermes ElevenLabs voice?
No. The voice is the built-in TTS provider. The sound effects plugin is a separate install. They can share the same API key, but a restricted key needs the Sound Effects permission too.
Can I use it in more than one Hermes profile?
Yes, but plugins install per profile. Run the install with the -p flag and the profile name for each one.
Ready to give your Hermes Agent ElevenLabs sound effects and a voice? Join the AI Profit Boardroom for the Agent OS and daily Hermes tutorials, or book a free AI strategy session and we'll build your setup together.
