Cartesia Sonic 3.6 heteronyms: read, bass, live, and Sume pricing

Cartesia lists higher pronunciation accuracy on English heteronyms such as "read", "bass" and "live" in context as a Sonic 3.6 change, so the first fix is a sentence with enough context around the word. The changelog snapshot read here gives no Sonic 3.6 price. On Sume, text to speech is listed at $0.0475 per 1,000 characters on the rate card, and a pronunciation_dict_id is available when context is not enough.

Sources: the Cartesia 2026 changelog and Sume's API reference, read 2026-10-01.

Is there a Sonic 3.6 price on the changelog?

No. The page describes behavior and migration, not rates, so this post states no Cartesia price. Check the vendor's pricing page for that.

When is context enough for a heteronym?

Put the clue next to the word. If a one-word line like "Read." must be said a certain way, context cannot help, and that is when an override is worth it.

When do I reach for a dictionary id?

Sume's request takes an optional pronunciation_dict_id, up to 128 characters. The cited schema does not describe building a dictionary or say it resolves heteronyms by sentence, so read the pronunciation post and test before relying on it.

What does Sume cost for a script?

The rate card lists text to speech at $0.0475 per 1,000 characters, and the request schema notes that spaces and punctuation count toward usage, with a 20000-character cap on transcript. Add the Router model such as sonic-3.6 when you call /v1/tts-router/generate.