Languages
Voices
The voices your customers hear on calls, how they are chosen and tested with your marketing team, and how a custom voice for your brand is built.
On this page
On a call, the voice is the first thing a customer hears of your brand. A luxury developer and a budget airline should not sound the same, and neither should sound like a generic system. Each deployment has its own voices: a choice for each language, male or female, with different characters, from voice synthesis providers running in your region. Your marketing team hears the shortlist, chooses, and approves the result before any customer hears it.
What you choose
Voice synthesis (the technology that speaks the agent's reply) offers several voices in each language. You choose one for each language your agent speaks.
| Choice | What it means | Example |
|---|---|---|
| Language | A voice for each language the agent speaks | An Arabic voice and an English voice with a similar character |
| Male or female | Both are available in each language | A female voice for guest services, a male voice for a claims line |
| Character | How the voice comes across: calm, warm, bright, measured | A calm, formal voice for a luxury developer. A bright, brisk voice for a budget airline |
| Arabic variety | The Arabic voice speaks the variety of Arabic agreed for your deployment | Gulf Arabic for a Riyadh hotel group's reservation calls |
| Custom voice | A voice unique to your brand, built from a recording of a speaker you choose | The voice of the presenter in your brand campaigns |
Different agents in your deployment can use different voices, for example one for sales enquiries and another for the claims line.
How voices are chosen and tested with your team
Choosing a voice is a brand decision, so your marketing team makes it, on your own material rather than generic samples.
- Brief. Your marketing team describes the character your brand should have in each language. Telonic agrees the brief with them, alongside the agent's tone.
- Shortlist. Telonic shortlists voices that fit the brief and handle your customers' dialect, from voice synthesis providers running in your region.
- Listen. Each shortlisted voice reads the same lines, written from your conversations: a greeting, an amount, a date, a booking or unit reference, your brand and project names, an apology and a handover.
- Correct. The pronunciation of your brand, project and place names, and of English terms inside Arabic sentences, is checked and corrected.
- Test. Your team calls the agent with the chosen voice on the real line before go-live (the point at which the agent starts handling real customers).
- Approve. Your team approves the voice for each language before it is released.
Custom voices for your brand
Where you want a voice that belongs to your brand alone, a custom voice is configured during implementation. It is built from a short recording of a speaker you choose, such as the presenter from your brand campaigns or a voice artist you commission.
The speaker gives written consent before the recording is made, covering how their voice will be used. The custom voice is used only in your deployment. If the speaker withdraws consent, or your agreement ends, the custom voice is taken out of use and deleted, with written confirmation.
Whatever voice it speaks in, including a custom voice, the agent tells customers it is an AI agent and offers a person whenever they ask for one.
How voice and tone work together
Tone decides the words: how formal, warm or brief the agent is. The voice decides how those words sound. The two are agreed together, and replies on calls are written for speaking, in short sentences, so the voice can deliver them naturally. See Tone, voice and brand.
Changing a voice after go-live
You can change a voice at any time, for example for a new brand campaign or a new line. A change of voice goes through the same steps as the first choice: listening on your own lines, testing on the real line, and approval by your team before it is released. See How changes go live.
The technical detail
| Question | Answer |
|---|---|
| Where does voice synthesis run? | Always in your chosen region. No voice synthesis provider outside the region is offered |
| How are providers chosen? | Per deployment, with you, based on the voices available, how natural they sound in your languages and dialects, and your region. Which providers and models are available in your region is confirmed during implementation, and named in your agreement. See Models and providers |
| Are amounts and dates read out correctly? | Figures such as amounts and dates are inserted directly from the record in your system, not retyped by the model, and the review lines check how the voice reads them |
| Is the voice provider a sub-processor? | Yes. The provider for your deployment is named in your agreement and in the list of sub-processors (companies that process personal data on Telonic's behalf). See Sub-processors |
| Does a change of voice provider need my approval? | Yes. Your team approves any change of model or provider before release |
In practice
A hotel group, Qamar Stays, sets up voices for its reservations line.
- The marketing team asks for a voice that sounds like its front desk: warm, formal and unhurried, in Gulf Arabic and English.
- Telonic shortlists two Arabic and two English voices from providers in the UAE region. Each reads the same lines, including the names of the group's resorts.
- The team chooses a female voice in each language. The pronunciation of one resort name is corrected in both.
- The reservations team makes test calls, including a booking change and a complaint that goes to a person, and approves the voices.
- For a new campaign, the group later asks for a custom voice from the presenter of its brand films. The presenter gives written consent, records a short script, and the new voice goes through the same review and approval before release.
What your team controls
- The voice for each language and each agent.
- Whether to build a custom voice, and whose voice it is.
- The pronunciation of your brand, project and place names.
- Approval of every voice, and every change of voice, before release.