Changelog

Simba 3.5 is the recommended model

simba-3.5 speaks 48 languages under one model id, on every plan, up from the six that simba-3.0 covers. It is the model GET /v1/audio/models marks recommended from today.

• One request field. Pass "model": "simba-3.5" and set language to the language of your text, for example ja-JP. Any regional variant of a listed language is in scope. See Simba 3.5 languages.
• Every endpoint you use today. POST /v1/audio/speech, POST /v1/audio/stream, POST /v1/audio/stream/with-timestamps and the WebSocket at GET /v1/audio/stream/ws. POST /v1/audio/dialogue stays on simba-dialogue-1.0.
• Every voice. Each catalog voice speaks all 48 languages on simba-3.5, and so does a voice you cloned on a paid plan, with the same voice ID.
• The SSML you use today. <break>, <prosody rate> and <sub alias> are applied; pitch, volume, emphasis and emotion tags are accepted and not applied, as on simba-3.2 and simba-3.0.

Nothing changes unless you set it. A request that omits model still resolves to simba-3.0, and simba-3.0 and simba-3.2 stay available and are not deprecated. No API version change is needed.

simba-multilingual is served by Simba 3.5 from 2026-11-21, on every API version that can still name it. Switching to simba-3.5 before then lets you pick the voice and test it first. See Simba 1.6 and your API version.

The reference is Models.

Say names, brands and terms your way with pronunciations

POST /v1/audio/speech, POST /v1/audio/stream and POST /v1/audio/stream/with-timestamps accept an optional pronunciations list on Simba 3.5 (simba-3.5) in every language it speaks except English, which is not supported yet, and the GET /v1/audio/stream/ws input stream takes the same list in a session.pronunciations message. Each entry names a word or phrase in your input and either the IPA phones to say (ipa) or the text to say in its place (alias), and applies everywhere that text appears.

• Matching. Entries ignore capitalization unless case_sensitive is true, match whole words (anywhere in Chinese, Japanese, Korean, Thai, Lao, Khmer and Burmese text), and the longest match wins. An entry overrides the model’s own pronunciation.
• Limits. Up to 200 entries; text up to 64 characters and one word for an ipa entry, ipa up to 64 phones from the model’s phone set, alias up to 160 characters, and the input with every alias in place up to 20,000 characters.
• Not billed. A request is billed for its input exactly as before, and speech marks still index your input.
• GET /v1/audio/models lists the languages each model takes them in as pronunciation_languages.
• A model or language without them returns 400 pronunciations_unsupported instead of ignoring the list. A malformed entry or a phone outside the phone set returns 400 validation_failed, naming the entry and the phone.

Omitting the field changes nothing. No API version change is needed. See Pronunciation.