Last updated
What is Gladia used for in a Bolna voice agent?
Gladia transcribes the caller's speech in real time for your Bolna agent, using its Solaria model with under 300ms of latency across more than 100 languages. It handles code-switching, so a caller who mixes English and Hindi is still understood. Bolna also turns on Gladia's audio enhancer for phone calls to help with background noise.
How do I set up Gladia with Bolna?
- Open your agent at platform.bolna.ai and go to the Languages tab.
- Under Transcription, set Provider to Gladia and choose the Solaria model.
- Set the caller's primary language. For Indian languages, Bolna also listens for English.
- Add company names and product terms in Keywords. They go to Gladia as custom vocabulary.
- On the Engine tab, adjust Endpointing to balance quick replies against callers who pause.
In the API, set provider to gladia and model to solaria-1 on tools_config.transcriber. See the Gladia setup guide in the Bolna docs.
What does Gladia support on Bolna?
| Feature | Detail |
|---|---|
| Model | Solaria (solaria-1), the default and recommended model |
| Latency | Under 300ms |
| Languages | 100+ |
| Code-switching | Primary language plus English, for markets like India |
| Keywords | Sent as custom vocabulary, no weights, no context field |
| Audio on Twilio | Native mulaw at 8kHz |
| Audio on Exotel and Plivo | linear16 at 8kHz |
| Audio on web calls | linear16 at 16kHz |
| Audio enhancer | On for Twilio, Exotel and Plivo calls |
| Endpointing | Configurable silence threshold |
What can you build with Gladia and Bolna?
Hinglish and mixed-language support lines. Gladia transcribes calls where people switch between an Indian language and English in the same conversation. For other code-switching options, compare Soniox.
Agents for European markets. Bolna's provider guide lists Gladia for European languages with good multilingual coverage.
Noisy phone lines. The audio enhancer runs on Twilio, Exotel and Plivo calls, and Twilio audio goes to Gladia as native mulaw with no conversion step.
Brand-heavy conversations. Custom vocabulary helps the agent catch company names, product names and industry terms that a general model would miss.