Skip to main content

Settings

The Settings tab controls how your voice agent sounds and behaves. Settings are organized into sections: Voice Settings, Call Settings, Post-Call Analysis, and Widget Settings.


Voice Settings

Language

Set the primary language for the agent's greeting using the language dropdown.

AspectDescription
Initial languageThe agent greets callers in this language
Auto-switchingThe agent automatically switches to the caller's language
Multi-languageThe agent understands and responds in multiple languages

The language setting primarily affects the greeting. Once the conversation begins, the agent adapts to whatever language the caller uses.

STT Languages

Use the multi-select field to specify up to 4 languages for speech-to-text recognition (e.g., en, de, ru, or the longer en-US, de-DE forms). This improves recognition accuracy for multilingual environments — for example, a German business that also serves English and Turkish speakers.

When set, only these languages are used for speech recognition. Leaving the field empty means the agent auto-detects based on the primary language.

Short vs long codes

Both short language codes (en, de) and region-specific codes (en-US, de-DE) work. The platform automatically picks a sensible regional variant when you give it a short code.

Voice Selection

Click the Voice field to open a selection modal where you can browse and preview available voices. Voices differ by gender, tone, accent, and speaking style.

Voice Fine-Tuning

SettingDescriptionRangeDefault
VolumeAdjust how loud the agent speaks (in dB)-90 to +1000
SpeedAdjust how fast the agent speaks-20 to +300

Choosing the Right Voice

Business TypeRecommended Voice Style
Professional servicesCalm, mature voice
Youth brandsEnergetic, younger voice
HealthcareWarm, reassuring voice
Tech supportClear, patient voice
SalesConfident, enthusiastic voice
tip

Preview voices before selecting — click on a voice name to hear a sample. Match the voice to your brand personality.

Background Ambience

(v2.8) Give the call a subtle sense of place by playing a background sound behind the agent's voice. Choose from a small set of ambiences, set how loud it should be, and preview it instantly before you commit.

AmbienceFeels like
OfficeA busy workplace behind the agent
Coffee shopRelaxed café atmosphere
Call centerA staffed support desk
OutdoorsOpen-air, outside
RainCalm, rainy background
ReceptionA front-desk / lobby feel
SettingDescriptionDefault
AmbienceWhich background sound to play (or none)Off
VolumeHow loud the ambience sits under the voice

Background Ambience is off by default. Use the instant preview to check the mix — the ambience should stay well under the agent's voice so it never competes with what the caller needs to hear.

When to use it

A light ambience can make an agent feel more human and less "recorded". Keep the volume low — a background that draws attention away from the conversation does more harm than good.

Region-Aware Localization

(v2.2) When you deploy an agent to a specific region — de-AT, de-CH, en-GB, and so on — the platform automatically applies that market's conventions at the start of the call. You don't configure anything; picking the region does the work.

This affects how the agent:

  • Reads numbers and currency in the local convention
  • Speaks banking codes (IBAN, sort codes, etc.) the way that market expects
  • Formats addresses to the local standard
  • Uses the region's TTS conventions so the voice sounds native rather than a generic variant of the language

The result: an de-AT agent sounds Austrian, an en-GB agent sounds British, without you tuning the prompt for each market.


Call Settings

End Call After Silence

A toggle that controls how the agent handles caller silence. When turned on, the call ends automatically after a period of silence. Use the slider to set the silence duration.

SettingDescriptionRangeDefault
ToggleEnable or disable silence-based call endingOn / OffOff
TimeoutSeconds of silence before ending the call10–30 seconds10s

When silence mode is on, the Reminder settings below are disabled — the two modes are mutually exclusive.

Reminder System

When silence mode is off, the agent uses reminders to re-engage silent callers instead of ending the call.

SettingDescriptionRangeDefault
Reminder TriggerSeconds of silence before sending a reminder10–30 seconds30s
Reminder Max CountHow many reminders before ending the call1–103

The reminder system works as follows:

  1. Caller goes silent
  2. After Reminder Trigger seconds, the agent sends a reminder (e.g., "Are you still there?")
  3. If still silent, repeats up to Reminder Max Count times
  4. After all reminders, the agent ends the call politely
tip

Either use End Call After Silence or Reminders — they are mutually exclusive. Silence detection ends the call immediately; reminders give the caller more chances to respond.

Interruption Sensitivity

Control how readily the agent stops talking when the caller starts to speak. This is a per-agent setting that sits alongside the silence and reminder controls.

SettingDescription
Interruption sensitivityHow eagerly the agent yields the floor when the caller speaks over it
  • Higher sensitivity — the agent stops almost immediately at the first sound from the caller. Good for fast, back-and-forth conversations, but background noise can cut the agent off mid-sentence.
  • Lower sensitivity — the agent needs a clearer, more sustained interruption before it stops. Good for noisy environments or agents that read out longer passages, but the caller has to be more deliberate to break in.
tip

If callers complain the agent talks over them, raise the sensitivity. If the agent keeps getting cut off by background noise, lower it.

Max Call Duration

Set a hard ceiling on call length. When the cap is reached, the agent politely ends the call no matter what's happening in the conversation. Useful for cost control on outbound campaigns, demo-call scenarios, and any setup where you want a guaranteed upper bound on talk time.

SettingDescriptionRangeDefault
ToggleEnable or disable the duration capOn / OffOff
DurationMaximum call length in minutes1–1510 (when first enabled)

The cap applies to every entry point for this agent — incoming phone calls, outbound dials, browser test calls, and callback widget calls. It does not affect calls handled by other agents.

When the cap fires, the agent ends the call within a couple of seconds. The call appears in the call log, flagged as ended by the duration limit.

When to use it

Leave the cap off for normal customer-support and receptionist agents — natural call length serves the conversation. Turn the cap on for outbound campaigns where unbounded calls inflate your costs, for demo agents you embed in landing pages, or for any agent talking to anonymous visitors.

Callback widget interaction

The Callback Widget has its own safety ceiling of 10 minutes per call. If you also set a Max Call Duration on the agent, the shorter of the two wins.

Customer Memory

Toggle to let the agent remember previous conversations with the same caller.

SettingDescriptionDefault
Customer MemoryToggle on/off — when enabled, the agent retains context from past calls with the same contactOff

When enabled:

  • The agent recognizes returning callers and references previous conversations
  • Context from past calls is available to the AI during the conversation
  • Creates a more personalized experience for repeat callers

Example: A returning customer calls about their order. The agent says: "Welcome back! Last time we spoke about your delivery to Vienna. How can I help today?"

Recording Enabled

Toggle call recording on or off. When enabled, all calls handled by this agent are recorded.

SettingDescriptionDefault
Recording EnabledToggle on/off — when enabled, all calls are recordedOff

When enabled:

  • Full audio recordings are saved with each call log
  • Access recordings from the Call Logs section
  • Useful for quality assurance, training, and compliance
Legal Compliance

Check local laws regarding call recording consent. Many jurisdictions require you to inform callers that they are being recorded. Consider adding a notice in your agent's first message.


Post-Call Analysis

Automatically analyze calls after they end.

Sentiment Analysis

Toggle to track the caller's emotional tone throughout the conversation.

SentimentDescription
PositiveCaller is satisfied, happy, or appreciative
NeutralNormal conversation tone
NegativeCaller is frustrated, unhappy, or angry

Enable this to:

  • Monitor customer satisfaction trends
  • Identify problematic call patterns
  • Improve agent prompts based on sentiment data
  • View sentiment in call logs

Call Summary

Toggle to automatically generate a summary of each call with key phrases and topics discussed.

When enabled, after each call you get:

  • A concise summary of the conversation
  • Key phrases and topics identified
  • Action items mentioned during the call

View summaries in the Call Logs section.


Quality & Testing Aids

Two built-in aids help you keep an agent on track — they live with the prompt on the Model tab rather than in Settings:

  • Prompt Tester — simulate a conversation with the agent right in the dashboard, then get concrete suggestions for improving the prompt based on how the simulated call went. Use it to catch gaps before real callers hit them.
  • Guardrails — monitor the agent for drift, flagging when its behaviour strays from what the prompt intends.

See Prompt Engineering for how to use these while iterating on your prompt.


Widget Settings

The agent's Widgets tab is the home for everything widget-related: the theme picker, the callback widget configuration with its own language picker, the embed snippets for the in-browser widget types, and the domain restrictions that apply to all widgets.

SettingDescriptionDefaultLimits
ThemeColor theme used by all widgets"tangerine"11 themes available
Callback widgetEnable the phone-callback widget, configure retry attempts, retry interval, and widget languageDisabledSee Callback Widget
Terms & ConditionsShow a consent dialog before the conversation startsDisabledContent max 5,000 chars
Allow all domainsAllow widget to be embedded on any websitetrue
Allowed domainsIf "Allow all" is off, the list of domains permitted to embed[]Max 50 domains

See Website Widgets for the in-browser widget types and embed snippets, and the dedicated Callback Widget page for the phone-callback widget.


Data Retrieval Field Types

When defining retrieval variables in the Actions tab, each field has a type:

TypeUse CaseExample
TextNames, addresses, notes, free-form inputCustomer name, delivery address
NumberQuantities, budgets, IDsOrder quantity, budget amount
EmailEmail addresses with validationCustomer email
PhonePhone numbers with validationCustomer phone number
SelectorChoice from predefined optionsPreferred plan (Basic/Pro/Enterprise)
CheckboxYes/no consent or confirmation"I agree to receive marketing emails"

See Actions & Tools for full details on configuring retrieval variables.


Settings Impact on Plans

SettingFreeStarterProBusinessEnterprise
Language selection
Voice selection
Silence/Reminder settings
Max Call Duration
Call recording
Sentiment analysis
Call summary

All settings are available on all plans. The Callback Widget (configured in the Widgets tab) requires a paid plan because it places outbound calls.