Audio / Text-to-Speech

Bulbul V3

Text-to-speech model from Sarvam AI · released Feb 2026 · vendor-reported (not yet ranked on an independent leaderboard)

Unrankednot on an independent leaderboard yet
Feb 2026release

Specifications

Languages
11 (BCP-47 codes, incl. Malayalam and Odia)
Voices
30+ named speakers
Input limit
2,500 characters per REST request
Launch pricing (reported)
₹30 per 10,000 characters
API model id
bulbul:v3
Lifecycle
End of life on Sarvam Voice Agents on 10 Nov 2026; remains the Model API default when `model` is omitted

Notes

  • Not ranked on the Artificial Analysis Speech Arena at the time of writing.

How this is measured

Speech Arena (Text-to-Speech)

Elo rating (higher is better)

Models are ranked with an Elo system derived from blind user votes: listeners hear two speech samples generated from the same text and pick the one that sounds more natural. The provider-voice leaderboard uses each provider's own native voices. Each result is reported with a 95% interval and the number of votes (samples).

Release & sources

Release
Feb 2026 (month precision)
Release source
Sarvam changelog (promoted from bulbul:v3-beta to stable, February 2026)

More TTS models from Sarvam AI