Launched this week
ElevenLabs Realtime Skill
An agent skill to build on ElevenLabs Realtime voice model
2 followers
An agent skill to build on ElevenLabs Realtime voice model
2 followers
The ElevenLabs Realtime skill teaches AI coding assistants to build on ElevenLabs Realtime models. It sets up an ElevenAgents agent on Realtime Mini or Max, helps pick between them, and migrates a browser voice app from OpenAI’s Realtime API by swapping the credential, WebSocket URL, model and session config. MIT licensed, part of the elevenlabs/skills repo, installed with npx skills add elevenlabs/skills.
The ElevenLabs Realtime skill teaches AI coding assistants how to build on ElevenLabs Realtime models, so you can set up a voice agent, or move an existing voice app over, without digging through the docs yourself.
Problem: Realtime voice models are new, and getting them working takes a lot of small details: which model to pick, which settings an agent ignores, which agents can’t use them, and how an app built on OpenAI’s Realtime API has to change. A coding assistant can easily get those wrong without the right instructions.
Solution: An agent skill in the open-source elevenlabs/skills repo. Your coding assistant loads it and follows ElevenLabs’ own steps for two jobs: running an ElevenAgents agent on a Realtime model, and migrating a voice app from OpenAI’s Realtime API.
What makes it different: It’s written by ElevenLabs for coding assistants, and it checks for problems before you hit them. For agents, it looks for blockers like a custom LLM, workflow overrides, text-only agents and transfers between modes, and it helps you choose between Mini and Max, including data residency and compliance limits. For migrations, it keeps the OpenAI Realtime SDK and changes where the credential is minted, the WebSocket URL, the model and the session config.
Key features:
Set an agent’s Voice mode to Realtime with conversation_config.realtime_model, using the CLI or the Python and TypeScript SDKs
Choose between Eleven Realtime v1 Mini (lowest latency and cost, the default) and v1 Max (reasons before it responds, for complex instructions and multi-step tool use)
Migrate a browser voice app from OpenAI’s Realtime API to the OpenAI-compatible ElevenLabs Realtime endpoint
Steps for wiring up audio after leaving WebRTC, since ElevenLabs Realtime is WebSocket-only
A table of the model and data residency options
Error messages with their causes and fixes
Who it’s for: Developers building voice agents who use a coding assistant, and teams with an OpenAI Realtime voice app who want to try ElevenLabs.
Worth knowing: It’s for browser apps, and it has no steps for server-side connections like relays or telephony bridges. Realtime models are rolling out gradually, so some workspaces don’t have access yet. You need an ElevenLabs API key, and Realtime usage is billed per ElevenAgents rates, with Max costing more than Mini.
Install it with npx skills add elevenlabs/skills, and see the Realtime skill on GitHub.