Sarvam AI charges ₹30 per hour for speech-to-text across all three processing modes: real-time, streaming, and batch. The billing unit is per second of audio, rounded up to the nearest second in each request.

For contact center forecasting, the practical implication is straightforward: convert your total call minutes to hours and apply the hourly rate. Ten hours of processed audio per day costs ₹300 at the base rate. The per-second rounding means short calls (under one minute) are charged for their exact duration, not a full minute minimum.

Billing is consolidated at the organization level. One credit balance covers every workspace, API key, and Sarvam product, which simplifies cost tracking across teams.

Free credits on signup

Every user receives ₹100 in free credits on signup, usable across any Sarvam API. The credits are universal and never expire, so they serve equally well for initial prototyping and for occasional low-volume testing in production environments.

At ₹30/hour, the free tier covers approximately two hours of base STT processing before any paid credits are required.

Feature-based pricing: diarization and translation

The configuration you choose has a direct effect on the per-hour rate. Sarvam publishes four STT variants:

Enterprise plans and volume discounts

Sarvam describes its enterprise tier as offering "tailored limits, throughput, pricing, and support for high-volume or production-critical workloads," including volume discounts on credits. Exact enterprise STT rates are not published publicly.

Teams running sustained contact center workloads should submit a request through Sarvam's enterprise contact page to obtain throughput guarantees and credit pricing that reflects actual monthly consumption.

Official source

The complete rate card, billing mechanics, and enterprise contact information are available at API Pricing.