Choose a voice and model
Pick how your agent listens, thinks and sounds.
Every voice agent uses three parts. Chat agents use only the model.
| Part | What it does |
|---|---|
| Transcriber | Turns what the caller says into text. |
| Model | The AI that understands the conversation and decides what to say. |
| Voice | Turns the agent's reply into speech. |
Telik picks a working combination when you create an agent, so you don't have to change anything to get started.
Change the voice, model or transcriber
Open the agent and select the Voice & Models tab (called Model for chat-only agents).
Change the part you want:
- Voice: filter by gender, accent or style, or search by name. Play a sample before you choose.
- Model: choose a provider, then a model.
- Transcriber: choose a provider and model that support your agent's languages.
Select Save agent, then test your agent to hear the difference.
Languages
Your agent's languages are set in the General tab. The first language is the agent's main language.
Only options that support your agent's languages are shown. If Telik says no voice or transcriber is available for a language, choose another language or contact support@telik.ai.
To stop the agent replying to speech in a language it doesn't support, turn on Only accept the agent's language in the Voice & Models tab.
Advanced voice settings
Most agents work well with the defaults. If callers find the agent interrupts them or is slow to reply, open Advanced settings in the Voice & Models tab. There you can control:
- how Telik decides the caller has finished speaking,
- what has to happen before the caller can interrupt the agent,
- how background noise is reduced.
Cost
Where pricing applies, the editor shows the price of each option, per minute for voice and per message for chat. Larger, more capable models usually cost more. A smaller model is often enough for straightforward jobs like answering FAQs or taking bookings.