ModelCards

Sonic 3.6

Audio modelCard summarized

Our fastest, most natural text-to-speech model, with native support for 44 languages.

Source: Introducing Sonic-3.6 (launch post), Sonic 3.6 docs page and August 2026 changelog (no model card published)

Use it via API

$0.065 / 1K chars
Code exampleCompare

Where to call it

  • Cartesia APIsonic-3.6

Supports

No feature list in the maker's docs.

Full feature support

Good for

  • Voice agents and production calls at scale
  • Confirmation codes, alphanumerics and heteronyms read correctly without preprocessing
  • Multilingual and Indic voice apps, including Hinglish in Devanagari or Latin script

Watch out for

Overview

Cartesia · 01TTS, front

Sonic 3.6

Our fastest, most natural text-to-speech model, with native support for 44 languages.

Typical price
$0.065 / 1K chars
Pro plan overage
Max length
—
Languages
44 languages, 61 locales
Modes
text-to-speech, streaming (bytes, SSE, WebSocket)
Released
27 Aug 2026

Sonic 3.6 is Cartesia's text-to-speech model, an update to Sonic 3.5 that is fully backwards compatible with it. Cartesia says it adapts pacing, intonation and emotion to the transcript without SSML or instructions, sounds native across 44 languages, adds Odia and Urdu, improves Hindi/Hinglish transcript following, and replies in under 90 ms. launch post; docs; changelog Aug 2026

sonic-3.6
Typical price
$0.065 / 1K chars
Languages
44 languages, 61 locales
Weights
Closed
Released
27 Aug 2026
  • text-to-speech
  • streaming (bytes, SSE, WebSocket)
  • input streaming (continuations)
  • voice cloning (instant and professional)
  • multilingual voices
  • speed, volume and emotion control
  • word and phoneme timestamps
  • infill