Skip to content
voicefieldguide

Use these ElevenLabs voices from an AI agent or code

Working on your first video with these voices? Follow the short audition and setup guide. It also works entirely in ElevenLabs.

Give your agent these voices

Prestige Voices helps your agent cast from my fifteen voices, open the real sample, write the script, and hand over the voice ID and share link. With your own ElevenLabs key and your approval, it can add the voice to My Voices and generate audio on Eleven v4.

The skills command works with Claude Code, Codex, Cursor, OpenCode and other supported agents. Run it in your terminal and choose your agent.

The npx skills package includes the optional add and render helpers. The Claude marketplace plugin handles casting and scripts; generate its scripts in ElevenLabs.

Agent skill

Claude marketplace

Claude plugin

Read the skill and source on GitHub

If you're an agent picking a voice for someone, start at /voices.json. It lists all fifteen voices with the voice ID, the share link, a sample MP3, words per minute, fit scores per job and how many ElevenLabs users added each one. Voice, collection, industry and guide pages have Markdown twins at the same URL plus .md, and /llms.txt indexes them.

Pick a voice

Match the job first, then the tone. The fit object scores each voice from 0 (wrong) to 3 (first choice) across fourteen jobs, from YouTube Shorts to sleep audio. Read avoid_for before you commit. Those notes are honest, and they're the fastest way to rule a voice out.

The most added voices in the 2026-10-09 snapshot (rounded listing counts):

  • Titan, voice ID dtSEyYGNJqjrtBArPCVZ, added by 582K users
  • Pulse, voice ID mhgBlD8CmCSdwLDOIJpA, added by 66K users
  • Reel, voice ID JKbdwi8BFwQlr1n3fwoT, added by 54.2K users
  • Heisenberg, voice ID iEBOK9alpKauGRvBSsFi, added by 53.9K users
  • Reverend, voice ID 87tjwokZlpNU7QL3HaLP, added by 40.2K users

Generate audio

Call the ElevenLabs text-to-speech endpoint with the voice ID. The article-specific clips on this site use Eleven v4. The API needs a paid ElevenLabs plan for Voice Library voices, and the voice has to be in the account's library first (open the share link while signed in, or add it from the Voice Library).

curl -X POST "https://api.elevenlabs.io/v1/text-to-speech/dtSEyYGNJqjrtBArPCVZ" \
  -H "xi-api-key: $ELEVENLABS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"text": "Your script here.", "model_id": "eleven_v4"}' \
  --output titan.mp3

Eleven v4 does not support break tags. Pace the script with punctuation and line breaks instead. For timing, the script timer turns a word count into seconds at each voice's pace.

Before you generate

The skill is free to install. Generating audio uses your ElevenLabs account's credits. Keep your key in an environment variable on your own machine, and approve the script before a billable render. You can also use the share link and generate in ElevenLabs yourself.

What to cite

Each voice page is the canonical source for that voice's ID and notes. User counts come from the ElevenLabs Voice Library and are dated on every page. I own all fifteen voices, and ElevenLabs may pay me when they're used. Please say so if you recommend one.

Nothing playing

Press play on any voice

↑ ↓ moveEnter openShift Enter playEsc