Integrations
AssemblyAI
How AssemblyAI's speech models can be used in your deployment, always in your chosen region, and what they receive.
On this page
AssemblyAI is one of the speech providers Telonic works with. Speech providers supply the models behind every spoken conversation: speech recognition (turning a caller's speech into text as they speak) and voice synthesis (speaking the agent's reply in the voice chosen for your brand). AssemblyAI provides speech recognition, voice synthesis, or both. Which of its models your deployment uses, and for which role, is decided with you during implementation.
What a speech provider does in your deployment
| Role | What it means for you | Example |
|---|---|---|
| Speech recognition | Callers are understood as they speak, and the transcript builds during the call rather than after it | A caller's words transcribed as they speak on a voice call or in browser voice, and a WhatsApp voice note transcribed when it arrives |
| Voice synthesis | The agent's reply is spoken in the voice chosen for your brand, in each language | A calm, formal Arabic voice and a matching English voice |
| Custom voice | A voice unique to your brand, built from a short recording of a speaker you choose. Configured during implementation | A voice recorded by a presenter your marketing team selects |
How it connects
Speech recognition and voice synthesis always run in your chosen region. No speech or voice provider outside your region is offered, whatever you choose for the language model. Speech providers are chosen for your deployment on two things: their coverage of the languages and dialects your customers use, and the voices that suit your brand.
Telonic's contract with every provider prohibits using your data to train or improve their models, and zero data retention arrangements (the provider keeps no copy of what it processes) are used wherever the provider offers them. Any change of speech provider or model is tested on conversations drawn from your industry and needs your approval before release. If a provider has an outage, failover switches only to another approved provider in your chosen region, tested in advance and configured during implementation.
What the provider receives
| Data | Needed for | Default |
|---|---|---|
| Call audio, as the caller speaks | Speech recognition | Only if the provider is chosen for recognition. Processed in your region |
| The text of each spoken reply | Voice synthesis | Only if the provider is chosen for synthesis. Processed in your region |
| A recording of the speaker you choose | Building a custom voice | Only if you commission a custom voice, with the speaker's written consent |
| Access to your systems or the customer record | Not needed | Not given |
| Your data, for training the provider's models | Not permitted | Prohibited by contract |
Calls are recorded only once the caller has given consent, and nothing is stored without your permission. See Models and providers.
In practice
A Dubai property developer reviews its speech providers some months after its agent started handling real customers.
- Telonic proposes a change of speech provider, deployed in the developer's region. AssemblyAI is among the providers considered.
- The proposed provider is tested on conversations drawn from the property industry and compared with the one in use.
- The results are shared with the developer, whose team approves the change before release.
- The new provider is added to the developer's list of sub-processors (companies that process personal data on Telonic's behalf), and the agent's limits, memory and integrations stay exactly as they were.
Related
- Speech recognition and voiceIntegrations
- Models and providersAgents
- VoicesLanguages
- Supported languages and Arabic dialectsLanguages
- Hosting and data residencySecurity and data protection
Product names and logos are trademarks of their owners. Their mention shows systems Telonic connects to and does not imply partnership or endorsement.