More room to experiment. Realtime voice at ₹2 a minute.

A voice tutor. A practice interview. An appointment assistant that speaks the caller’s language. Each starts with the same question: can we afford to try this properly?

At ₹2 per connected minute, a 90-second session costs ₹3. One hundred three-minute sessions cost ₹600 in realtime usage. One thousand connected minutes cost ₹2,000. These are usage calculations, not estimates of customer behavior or promises about the total cost of running your business.

The developer API measures connected conversation duration. It does not round every conversation up to a whole minute. Final usage rounds up once to the nearest paise.

Before a session starts, the wallet reserves the maximum possible charge. A ten-minute maximum reserves ₹20. If the conversation ends after three minutes, the usage charge is ₹6 and the unused reservation is released. Set the limit to suit the experience you are building.

Choose one job that a conversation can complete. Let a user describe a need, ask a useful follow-up, and check whether the result helps them. Measure session duration, successful outcomes, and the places where people repeat themselves.

A lower per-minute price gives you room to iterate. A useful result is still the goal. A short conversation that fails the task is not a saving.

The integration preview lets you configure instructions, language, voice, reflex answers, and optional conversation steps. Request a session, wait for readiness, then connect your application. You do not need to create a saved agent for every experiment.

The ₹2 rate is for developer realtime conversation usage. Existing Cloud subscriptions remain separate. Standalone STT and TTS are coming soon; their prices are not part of this announcement.

Try the voice playground and read the developer documentation to plan your integration. Public paid API access is still in preparation. We want builders to understand the interface and economics before they commit to a rollout.