AI Co-Host

Automate → AI Co-Host configures when the AI responds, what it sounds like, and which chat prompts can wake it. The pill in the top-right corner shows what it’s doing, for example Speaking to the audience.

The AI Co-Host page: the co-host running, Ollama chosen as the provider at http://127.0.0.1, and ElevenLabs voices to speak with. Workspace and custom voice names are blurred.

Turn it on or off

The Co-host card has a single Running switch. Off means nothing on this page runs, whatever it’s set to.

Choose a model

Under Model, pick the Provider that writes the co-host’s replies: Ollama, OpenAI, Anthropic, Gemini or Other.

Run it on your own machine with Ollama

With Ollama, the co-host runs on your own machine; nothing leaves it unless the model you choose is a cloud one.

Install Ollama and download a model, for example:
Terminal window
```
ollama pull llama3.2
```
In ilyStream, check the Address. Ollama’s default is http://127.0.0.1:11434.
Press Find models and choose the model you downloaded.

Use a hosted provider

Choose OpenAI, Anthropic, Gemini or Other to use a hosted model. The chat prompts that wake the co-host are sent to that provider under its terms.

Try it before your audience does

Type into Try it the way chat would, and press Ask. The line underneath tells you whether it has answered.

Voice

Under Voice, choose how the co-host sounds when it speaks:

Same as chat — the voice Text-to-Speech uses.
Installed — a voice installed on your PC.
ElevenLabs — any voice from your saved ElevenLabs accounts. Search voices by name.