📌 Talk & listen
Two free upgrades make Hermes feel like a conversation, not a chat window: text-to-speech (it talks back, ElevenLabs) and speech-to-text (it hears you / transcribes audio, Groq). This page covers both.
AText-to-speech — make it talk
Free tier, real voices
ElevenLabs' free plan includes a monthly credit allowance (about 10,000 credits ≈ 10–20 minutes of audio, refreshed each month). Plenty to hear your agent talk and to try the premium voices.
Voice library
The library has hundreds of premade voices across styles and accents. Free accounts can use the standard premade voices; some high-end premium voices need a paid tier.
Zara & friends
Premium voices like the warm Zara style sound remarkably natural. If you love a voice, you can even clone a similar custom voice on paid plans.
Simple auth
ElevenLabs uses a single API key — no IAM, no cloud console. Hermes reads it from ELEVENLABS_API_KEY in your secrets file.
BSpeech-to-text — make it hear
Fast & free
Groq runs Whisper on ultra-fast LPU hardware. On Groq's free tier transcription is genuinely free — no card, generous limits.
Transcribe anything
Send a voice note, lecture recording, or interview clip and Hermes turns it into clean text it can summarize, organize, and file.
Multilingual
The whisper-large-v3-turbo model transcribes in many languages, not just English.
Simple auth
One API key again. Hermes reads GROQ_API_KEY from your secrets file.
💜 Meet Zara
Zara is a warm, natural-sounding female voice in ElevenLabs' library — the kind of voice that makes a text-to-speech assistant feel like a person. It's a favorite for a reason.
A few Zara voices exist in the library (e.g. "Zara – Warm", "Zara – Premium Customer Care", "Zara – Soft Spoken"). Pick whichever Zara sounds right, then grab its voice ID from the library or your Voices page — it's the short string under the voice's name.
🪜 How to set it up
From zero to a talking Hermes in a few minutes.
1Create a free ElevenLabs account elevenlabs.io
- Go to elevenlabs.io and click Sign up (Google or email both work).
- Choose the free (Starter) plan. You'll get your monthly credit allowance immediately.
- Browse Voices → Voice Library to preview options, and add Zara to your voices if you'd like to use it.
2Get your API key
- Open elevenlabs.io/app → Profile → API Keys (or API Keys in settings).
- Copy the API key (a long string). Keep it secret — anyone with it can use your credit.
3Grab Zara's voice ID
- Go to your Voices page (or the Voice Library) and open the Zara voice you chose.
- Copy its voice ID — a short alphanumeric string (not your API key) that appears under the voice's name.
4Wire it into Hermes
- On your server, add your key to Hermes' secrets file:
- Then point Hermes at ElevenLabs and set your voice. Replace YOUR_ZARA_VOICE_ID with the Zara id you copied in step 3:
5Hear it
- Restart Hermes (or start a new session) so it picks up the new provider.
- Ask it to speak: “read your answer aloud” or toggle voice mode. Hermes will respond in Zara's voice.
🧰 Troubleshooting
| Problem | Fix |
|---|---|
| No audio / still robotic | Confirm tts.provider is elevenlabs (hermes config get tts.provider) and the key is in ~/.hermes/.env, then start a fresh session. |
| “Invalid API key” | Re-copy the key from your ElevenLabs dashboard — make sure there's no trailing space/newline in .env. |
| Voice not found / error | The voice ID isn't valid for your account. Re-copy Zara's exact ID from your Voices page. |
| Out of credits | Free tier refreshes monthly. Until then, switch back to edge (hermes config set tts.provider edge) — it always works. |
| Premium voice locked | Some library voices need a paid plan. Use a free Zara-style voice, or upgrade if you use voice a lot. |
🗣️ Hear it: connect Groq for voice transcription
With a free Groq account, Hermes transcribes the voice notes, lecture recordings, and audio you send it — so you can hand it a voice memo and get back clean, organized text. Hermes uses Groq's whisper-large-v3-turbo model, which is free on the free tier (no card, generous daily limits).
1Create a free Groq account & key
- Go to https://console.groq.com and sign up free (email or Google — no credit card).
- Open API Keys and click Create API Key.
- Copy the key (it starts
gsk_) — keep it secret, like a password.
2Add the key to Hermes
Secrets live in ~/.hermes/.env on your server. Add your Groq key there:
3Tell Hermes to use Groq for transcription
- Point Hermes' speech-to-text provider at Groq:
The default language is English. If you'll send non-English audio, set your language once: hermes config set stt.language en (swap en for e.g. es, zh, ...).
4Use it
- Start a new Hermes session (so it picks up the new provider).
- In Telegram, send the bot a voice message, or in chat attach an audio file and ask "transcribe this and summarize it."
- Hermes turns it into text it can then summarize, file, or act on.