All posts
Engineering
Measuring turn-taking in live calls
How our team evaluates when an agent should speak, wait or yield, and the tools we use to test it before launch.
Sam Reyes
6 min read
Knowing when to talk is as important as knowing what to say. An agent that interrupts feels rude, and one that waits too long feels broken.
Signals we listen for
Quill streams partial transcripts along with cues such as falling pitch and short pauses. The agent combines them to decide whether the caller has finished a thought.
Testing before launch
We replay recorded conversations against every new release and score how often the agent cuts in, misses a turn or talks over a caller. Regressions block the release.
Tuning for your callers
Every deployment can adjust how patient the agent is. A clinic line might wait longer, while a busy order line keeps things brisk.
More from the blog
News
Introducing Halvora Agents
Halvora Agents brings our speech models together into one platform for building phone agents that answer, route and resolve calls.
Mara Ellison
Research
How streaming state space models keep voice fast
A look at why we build our speech models on streaming sequence architectures, and what that means for live conversation.
Ivo Petrak
Product
Designing voices callers trust
What we learned while designing the default voice library, from pacing and warmth to how a voice should sound when it apologizes.
Lena Okafor
Get started today
Talk to an expert.
Meet someone from our team and see how Halvora can power voice experiences your callers will remember.
Start building.
Spin up a voice agent or call our models from the API, and go live the same day.
We’re hiring!
Help build voice that feels human