Last updated
What is Azure Speech-to-Text used for in a Bolna voice agent?
Azure Speech-to-Text converts the caller's speech into text in real time so your Bolna agent's LLM can reply, using Microsoft's streaming speech recognition. Bolna's provider guide recommends it for enterprise teams and agents that need many languages. It also adds speaker diarization and language identification for multi-party or multilingual calls.
How do I set up Azure Speech-to-Text with Bolna?
- Open your agent at platform.bolna.ai and go to the Languages tab.
- Under Transcription, set Provider to Azure and Model to
azure. - Set the language your callers speak.
- Add brand names and jargon in Keywords. Azure takes up to 2,000 phrases. The Context box stays disabled for Azure.
- To bill Azure usage to your own account, connect it at platform.bolna.ai/auth/azure.
In the API, set provider to azure and model to azure on tools_config.transcriber. See the Azure setup guide in the Bolna docs.
What does Azure speech-to-text support on Bolna?
| Feature | Detail |
|---|---|
| API values | provider: azure, model: azure |
| Streaming | Real-time transcription of live calls |
| Speaker diarization | Separates the agent and the caller in the transcript |
| Language identification | Detects languages in multilingual calls |
| Noise reduction | Noise suppression for difficult audio |
| Keywords | Phrase list, up to 2,000 phrases, no weights |
| Context field | Not supported |
| Pricing | On the preferred model list, included in the flat rate |
| Indian server configuration | Supported transcriber |
What can you build with Azure and Bolna?
Enterprise support lines. The docs position Azure for enterprise use and many languages. Pair it with Azure OpenAI and Azure text-to-speech to keep the whole agent on Microsoft.
Agents on Bolna's Indian servers. Azure is one of the transcribers allowed on the Indian server configuration, alongside Deepgram, Sarvam, ElevenLabs and Smallest.
Calls with heavy product vocabulary. Azure takes a phrase list of up to 2,000 phrases. Use it for catalogues, account terms or place names that a general model would miss.
Flat-rate, high-volume calling. Because azure sits on the preferred model list, its transcription is covered by Bolna's flat per-minute rate. See pricing.