Webinar Recap Building Ai Agents That Sound Natural
- Source
- elevenlabs.io
- Author
- ElevenLabs editorial sitemap
- Date
Context
A voice 's timing should fit what the caller is doing. The workshop recap recommends setting the agent's eagerness deliberately: an eager setting suits snappy, high-information exchanges, while a patient one suits someone who needs time to retrieve a document or number. Defaulting to eager across the board, it says, can make an agent feel pushy rather than natural.
The recap also distinguishes actual response time from perceived waiting. Starting speculative response generation before a caller finishes, or using brief filler speech during a delay, can make a pause feel shorter without shortening the underlying work. These are the presenter's recommended practices, not evidence of a measured improvement in customer outcomes.
- ElevenLabs' Luigi Sambuy names four traits of natural-sounding voice agents: rhythm and flow, mirroring the caller's emotion, from user history and business systems, and a brand-specific voice identity.
- In the restaurant demo, the natural behavior came from the rather than a different model: it permitted mid-sentence self-correction, trailing off while looking something up, and shifting tone to match the caller.
- Practical levers include emotional tags placed for the moment rather than applied uniformly, eagerness settings matched to the call (patient when a caller must find a document), pronunciation dictionaries, and deliberate background noise for callers used to contact centers.
- AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
- context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
- system prompt — The standing instructions a model receives before any user input — defining its role, rules, tools, and tone for the whole conversation.
Checking sign-in…
Loading comments…
