What is Gemini 3.5 Transcribe?
Gemini 3.5 Transcribe is Google’s real-time speech-to-text model, delivered over the Gemini Live API. It provides streaming automatic speech recognition (ASR) with high accuracy and low latency across a wide range of languages.
Bolna uses the gemini-3.5-transcribe-live model. It streams the caller’s words as they speak and produces a finalized transcript at each natural pause, so your agent can respond quickly.
Why choose Gemini for voice AI transcription?
-
High accuracy, low latency: Gemini streams interim results within a fraction of a second and finalizes each turn as soon as the caller pauses, keeping conversations responsive.
-
Broad language coverage: The model is trained across many languages, so you can run agents for a wide set of markets with a single provider.
-
Robust to mixed speech: When a caller drops in the occasional word from another language, the model handles it naturally rather than mis-hearing the turn.
-
Telephony and web ready: Works on both telephony (8 kHz) and web (16 kHz) calls with no extra configuration.
How Bolna uses Gemini for STT
Bolna connects to Gemini over a single streaming connection and biases it toward the language you select:
- Single language: Choose the language your callers speak (for example Hindi or English). Bolna hints that language so Gemini is tuned for it throughout the call, while still handling occasional words from another language naturally.
Interim results stream continuously so barge-in stays responsive, and each turn is finalized as soon as the caller stops speaking.
Which Gemini models are supported on Bolna AI?
Supported languages
Gemini on Bolna supports the following languages:
Set the language explicitly to the one your callers speak. Gemini is tuned to that language for the call and still handles the occasional word from another language without missing the turn.
Next steps
Ready to use Gemini for your voice AI agent? Open the Audio tab in the Bolna Playground, select Gemini 3.5 Transcribe as your transcriber, and choose your language.
For related integrations:
