Skip to main content

November 18th, 2026 Release

This release adds a new Sarvam voice model and makes the voice library easier to browse.

Voices & Text-to-Speech​

Sarvam Bulbul v4 Flash – bulbul:v4-flash is now available as a Sarvam text-to-speech model for your agents. If no voice is selected for this model, the agent falls back to a default Bulbul v4 speaker.

Grouped voice library – Voices that come in standard and Pro variants now appear together as a single voice in the voice library. Voices with only one variant are listed on their own.

Voice descriptions – Each voice now shows its own description in the voice library. In the agent voice picker, hover over a voice to see its description.

Cartesia default model – The default Cartesia text-to-speech model is back to sonic-3. Agents that have a Cartesia model selected explicitly are not affected.

See Text-to-Speech and Voice Selection for more on configuring agent voices.