How your agent sounds

The words matter, but so does the delivery. The same sentence can sound warm or robotic depending on the voice and the pacing. Here’s what you get to choose.

​

The voice

You pick a voice from a catalogue, the way you’d cast a person for a role. Each voice has its own sound and feel. Try a few out loud before you settle. The right voice depends on your audience and your brand, there’s no single “best” one. Gender and accent are properties of the voice you pick rather than separate switches. Asking for “a British female voice” narrows the catalogue to the voices that are one; you do not set accent independently afterwards.

​

Speaking speed

This is how fast the agent talks. Normal speed is 1. Lower is slower and calmer, higher is quicker.

• Slow down for older audiences, or anything with numbers and details.
• Speed up only a little, if at all. Fast talkers are hard to trust.

​

Expressiveness

This controls how much the voice varies its tone, its “expressiveness.” On screen it is Temperature, running from Steady to Expressive. A little variation sounds alive. Too much sounds theatrical. A middle setting is usually the safe, natural choice.

​

Language

You set the language the agent listens and speaks in, like US English. You can even give it a list of languages, and it will handle callers in more than one. Worth doing only if you genuinely get calls in both, because every extra language is a little more room for the agent to mishear. Over the API, language takes one locale or a list of them. The locales are en-US, en-IN, en-GB, en-AU, en-NZ, de-DE, es-ES, es-419, hi-IN, fr-FR, fr-CA, ja-JP, pt-PT, pt-BR, zh-CN, ru-RU, it-IT, ko-KR, nl-NL, nl-BE, pl-PL, tr-TR, vi-VN, ro-RO, bg-BG, ca-ES, th-TH, da-DK, fi-FI, el-GR, hu-HU, id-ID, no-NO, sk-SK, sv-SE, lt-LT, lv-LV, cs-CZ, ms-MY, af-ZA, ar-SA, az-AZ, bs-BA, cy-GB, fa-IR, fil-PH, gl-ES, he-IL, hr-HR, hy-AM, is-IS, kk-KZ, kn-IN, mk-MK, mr-IN, ne-NP, sl-SI, sr-RS, sw-KE, ta-IN, ur-IN, uk-UA. Four of them can only be used alone, not in a list: is-IS, fil-PH, hy-AM and ne-NP.

​

Background sound

With None picked the line is silent, and a perfectly silent line is one of the main reasons an agent sounds robotic. A light background, like Call center or Coffee shop with a low Volume, makes it feel like a real person on a real phone. If your agent sounds robotic or too clean, check this setting first.

​

Denoising mode

Real phone calls are noisy. Denoising mode cleans up what the agent hears before it works out what the caller said. Remove noise is the default and filters out background noise. Remove noise + background speech also drops other people talking near the caller, but can clip the caller’s own voice. No denoising passes the audio through unchanged. Stay on the default unless callers are often somewhere busy with other voices around them.

​

Do / Don’t

• Do listen to the agent out loud before going live. Your ear will catch things the settings page won’t.
• Do match the pace to your audience, slower for detail heavy calls.
• Don’t crank expressiveness to the top. Natural beats dramatic.
• Do keep a background sound quiet. It should be felt, not noticed.
• Don’t turn the background off to fix a robotic sound. Silence is what makes it sound robotic.

​

In the API

Every setting on this page is a field on the agent’s config, set with Update agent. Three more settings shape the conversation itself and live under config.interaction:

• enableBackchannel, backchannelFrequency and backchannelWords: the small “mm-hm” and “right” sounds the agent makes while the lead is talking, and how often.
• fillerWords: whether the agent uses natural fillers like “so” and “well”.
• responsiveness: how quickly the agent starts talking after the lead stops. Higher is snappier and interrupts more.

config.callScreening.callPurpose is the one-line reason the agent gives when a phone’s call-screening assistant asks why it is calling. In a split test, voice, language, interaction and callScreening are all per-variant, so you can test any of them.