Advance Settings — Build a Lively Caller Experience
Written By Anand Raj
Last updated 15 days ago
Advance Settings is where you tune how your AI agent feels on a call — how creative it sounds, what background it picks up, how it fills silence while thinking, how it pronounces tricky words, and how long it stays on the line. Good defaults work for most businesses, but these knobs let you shape the caller experience for your industry.
Think of the AI agent like a human brain. It may take a beat to look things up. It can mishear unclear speech. Accents shift from one region to another — and just like a human, the agent won't catch everything. Advance Settings exist to cover the maximum of those gaps.
Speech Settings
Answer Creativity
Controls how creative or precise the responses are. Increase it if the agent sounds too robotic or scripted.
Warning: Higher creativity can make the agent drift from the goal — it may start improvising from its general knowledge, which won't always make sense for your business. Move this up in small steps and test.
Speech Speed Sensitivity
Adjusts how quickly the assistant pauses when the caller interrupts. Lower it if the agent cuts the caller off, raise it if the agent waits too long before reacting.
Ambient Sound
Background sound played quietly under the call (e.g. Coffee Shop, office, café, or none). A small amount of ambient sound makes the call feel real and human instead of sterile. Use the volume slider to keep it subtle.
Thinking Sound
The AI sometimes needs a moment to read the right data from your systems. A Thinking Sound (e.g. Hold Music, soft tone, light typing) fills that gap so the caller doesn't hear dead silence and assume the line dropped.
Reminder Message, Pronunciation & Booster Keywords
Reminder Message Frequency
Controls how often the AI gently nudges the caller if they go quiet — set the Seconds to wait before reminding, and Times to limit how many reminders it sends before moving on.
Pronunciation
Every industry has words the model doesn't naturally pronounce well — brand names, medication names, locations, abbreviations. Click + Add and guide the model to say a specific word, name or phrase the right way.
Booster Keywords (Real-time Transcription)
Sometimes the AI mis-spells or mis-hears domain-specific words (your brand, product names, providers). Add them as Booster Keywords to expand the model's vocabulary so it picks up the right spelling on the call.
Call Settings
End Call Silence
If the caller stops speaking, this sets how long the agent waits before it ends the call automatically. Useful for catching dropped calls or callers who walked away, without holding the line forever.
Max Call Duration
Sets an upper cap on call length. Most calls resolve in 6–7 minutes — that's usually enough to answer a customer query or finish a workflow. If a call runs much longer, often the caller isn't sure what they need and the AI's quality starts to drop on very long conversations. A max duration also keeps your usage bill predictable.