Set a voice per language on a call bot
One agent, right accent. Voice lives on the bot — never on a campaign or a channel — and the tier you get depends on your plan.
Voice is set per language, on the bot. Not on the channel it answers, not on the campaign that dials — on the bot itself. This is the same principle as everything else about bots: the bot owns how it behaves, and the things pointing at it only decide when it is used.
The practical consequence is that one agent can sound right in more than one market without duplicating it. The other consequence is that if you need genuinely different voices for the same language in different campaigns, you need two bots.
Before you start
- A call bot, not a chat bot.
- Two or three candidate voices per language to audition. Every plan has the same ultra-realistic voices, so this is a taste decision rather than a plan one.
- Somebody who speaks each language you are enabling.
Steps
Open the call bot and find the voice settings
Only call bots have them. If you cannot find them, check the Type — you are probably looking at a chat bot.
Add only the languages you can support downstream
A bot that converses beautifully in Hindi and then hands over to a team who cannot has made the experience worse, not better. Enable a language when the handover works in it too.
Pick a voice per language
The same ultra-realistic voices are available on every plan, so nothing here depends on what you are paying. Pick on how the voice actually sounds saying your own opening line, not on tier.
Listen to each voice saying your actual first message
Not a generic sample. Your opening line, with your brand name in it. Brand names are exactly where synthetic voices most often break the spell, and you will hear it immediately.
Test the things you say every day
Prices, phone numbers, street names, appointment times. These are where pronunciation fails in ways that cost you bookings, and they are testable in about two minutes.
Fix pronunciation in the instruction
Spelling a difficult name phonetically in the instruction usually fixes it. This is the practical workaround and it works well enough that it is worth doing rather than accepting a mispronounced brand.
Confirm the switch mid-call
Have somebody change language partway through a call and listen to the seam. A clean switch is impressive; a bad one is the moment the caller realises they are talking to a machine.
Check it worked
Record one call per language and play it to somebody who speaks it natively. Ask them one question: would you keep listening? That answer is worth more than any setting.
The parts people get wrong
- A campaign cannot override the voice. If you need a different voice, you need a different bot.
- Because voice is per language on the bot, the same agent can serve two markets from one configuration.
The ones people actually ask
The voice sounds robotic even though everything is configured.
It is not your plan — the ultra-realistic voices are on every plan, so there is no lower tier to be stuck on. Try a different voice for that language, and listen to it saying your actual first message rather than a generic sample. If the pauses rather than the timbre are the problem, that is the model, not the voice.
The bot will not switch language when a caller does.
The language is not enabled on the bot. A bot will not switch into a language it has not been allowed to speak. Add it — but only if your handover works in that language too, because an agent that converses beautifully in Hindi and then hands to a team who cannot has made the experience worse.
It mispronounces our company name.
Spell it phonetically in the instruction. This is a practical workaround and it works well enough to be worth doing rather than accepting a mispronounced brand. Test it saying your actual first message, not a generic sample — brand names are exactly where synthetic voices break the spell.