Transcriber
Deepgram logo

Deepgram speech-to-text for Bolna AI voice agents

Deepgram is Bolna's default English transcriber, turning caller speech into text with Nova-3, Nova-2 or Flux streaming models on live phone calls.

stttranscriberdeepgramspeech-to-textvoice-recognition
At a glance

How Deepgram fits in the stack

Best for

Turning noisy phone audio into text the agent can reason over.

Use this layer when

Recognition accuracy, language coverage, and latency matter most.

Connects to

Telephony audio upstream and your LLM downstream.

Voice Stack

Speech-to-text is the listening layer

Speech-to-text is the listening layer

This provider converts raw audio into text in real time. It shapes how accurately the agent hears the caller and how natural the back-and-forth feels.

TelephonyPhone Network
→
STTListener
→
LLMReasoning
→
TTSVoice
→
ToolsActions

Last updated

What is Deepgram used for in a Bolna voice agent?

Deepgram is Bolna's default English transcriber: it converts the caller's speech to text in real time so your agent's LLM can respond, with the streaming nova-3 model as the recommended default. Pick it for English agents where latency matters most, or use nova-2 and its domain variants for other languages and specialist vocabulary.

How do I set up Deepgram with Bolna?

  1. Open your agent at platform.bolna.ai and go to the Languages tab.
  2. Under Transcription, set Provider to deepgram and Model to nova-3 for English, or nova-2 for other languages.
  3. Set the language explicitly. Auto-detect adds 100 to 200 ms and can misclassify.
  4. Add product names and jargon in Keywords. Keep it to 20 to 50 focused terms.
  5. To use your own Deepgram account, connect it at platform.bolna.ai/auth/deepgram.

For endpointing, encoding and every model, see the Deepgram setup guide in the Bolna docs.

Which Deepgram models does Bolna support?

ModelLanguagesBest for
nova-3EnglishLowest latency, recommended default
nova-3-medicalEnglishClinical terminology
nova-230+ languagesMultilingual agents
nova-2-phonecallEnglishPhone audio with background noise
nova-2-conversationalaiEnglishAI agent conversations
nova-2-meetingEnglishMultiple speakers
nova-2-medicalEnglishMedical terminology
nova-2-financeEnglishFinancial and trading terminology
nova-2-atcEnglishAir traffic control
nova-2-drivethruEnglishQuick-service ordering
nova-2-automotiveEnglishIn-car applications
Flux (English)EnglishTurn detection inside the model
Flux (Multilingual)Multiple, with per-turn language identificationMultilingual agents without a preset language

For Indian languages, the docs note Sarvam may give better accuracy.

What can you build with Deepgram and Bolna?

Fast English support lines. nova-3 is the lowest-latency option for English, which keeps dead air short on inbound and outbound calls. The same config works for both.

Healthcare and finance calls. Use nova-3-medical, nova-2-medical or nova-2-finance so clinical or trading terms come through correctly.

Snappier turn-taking with Flux. Flux can start the LLM response before the caller fully stops and detects barge-ins without VAD delay. See the Deepgram Flux guide.

Cost-sensitive, high-volume campaigns. The docs recommend Deepgram nova-2 for high-volume calling where cost matters, and it sits inside Bolna's flat per-minute rate. See pricing.

FAQ

Deepgram on Bolna: frequently asked questions

Use nova-3 for English agents, since it is faster and more accurate. Use nova-2 with the language set explicitly for non-English, or a domain variant like nova-2-phonecall or nova-2-medical.

Yes. Connect it at platform.bolna.ai/auth/deepgram or under Settings → Integrations. Bolna does not charge for usage on providers you connect yourself.

Yes. nova-2 and nova-3 are on Bolna's preferred model list, so they are covered by the flat rate of $0.06/min (₹5.52/min) at standard wallet tiers.

Your endpointing value is too low. The default is 250 ms; raise it to 350 to 500 ms for non-native speakers, elderly callers or anyone who pauses mid-sentence.

Browse this layer

Keep exploring the voice stack

Browse

Speech-to-Text

Speech-to-text converts what callers say into text that your LLM can process. Transcription accuracy and latency directly affect how natural a conversation feels. Bolna supports streaming STT providers optimized for telephony audio, including specialized models for Indian languages.

Browse

Telephony

Telephony providers connect your voice agents to the phone network so they can make and receive real calls. Bolna supports managed integrations with major carriers as well as bring-your-own-carrier via SIP trunking, giving you full control over call routing, number provisioning, and cost.

Browse

Large Language Models

The LLM is the brain of your voice agent. It understands what callers say and decides how to respond. Bolna lets you switch between models like GPT-5.4 mini, Claude and DeepSeek by changing the provider and model in your agent LLM settings, so you can tune for speed, cost or reasoning depth.

Browse

Text-to-Speech

Text-to-speech turns your agent responses into spoken audio. Voice quality shapes how callers perceive your brand. Flat, robotic speech kills trust while natural, expressive voices build it. Bolna integrates with the fastest TTS providers so responses sound human and arrive without awkward pauses.

Browse

Tools & Workflows

Tools let your voice agents take action during a call, not just talk. Book a calendar slot, look up an order through your own API, push a lead into your CRM, or trigger a multi-step automation in Zapier. These integrations turn voice agents from answering machines into workflow engines.

See where Deepgram fits in your production workflow

Use the demo to walk through provider selection, stack tradeoffs, and the exact workflow you want Bolna to automate.