Language Support
SpeechifyAI Build models support synthesizing speech in multiple languages. The API handles both single-language texts and mixed-language inputs.
Use simba-3.5 for any of its 48 languages, or simba-3.0 for its six. simba-multilingual needs a workspace pinned to an API version before 2026-09-21 and is served by Simba 3.5 from 2026-11-21. simba-3.2 and simba-english are English-only; a non-English voice on them returns 400.
Which model covers which languages
Simba 3.5 languages
simba-3.5 is streaming-native and speaks these 48 languages under one model id, on every plan. Send the code as the language parameter, either alone or with any region: pt, pt-BR and pt-PT are all Portuguese, and ja and ja-JP are both Japanese.
Every voice in the catalog speaks each of these languages on simba-3.5, your workspace’s cloned voices included. When you omit language, the voice’s own locale decides, so set it whenever the text is not in the voice’s language. A language outside this list is not rejected, but it is not validated either.
Staff picks for every language
Simba 3.5 sounds best with a voice native to the language you generate. Every voice keeps the accent of the recording it was made from, so an English voice reading German sounds like an English speaker speaking German. A cloned voice works the same way: clone from a recording in the language you will use it in.
We kept this on purpose. Customers asked that a voice, and above all their own clone, keeps sounding like itself in every language. We plan to make this more flexible in an upcoming release.
For each language, start with its top voices.
These are our staff picks, the voices we recommend first for each language.
In GET /v1/voices, each carries staff_pick_rank: 1 for the first pick of its locale, 2 for the next, and so on.
Filter with the full locale, such as de-DE, and start from the lowest staff_pick_rank.
The console marks the same voices as staff picks.
For a language without a row here, choose a voice whose locale is that language.
Simba 3.0 languages
simba-3.0 is streaming-native and officially supports six languages across these seven locales:
es-ES and es-MX voices are interchangeable on simba-3.0: pass language: es-ES with an es-MX-labeled voice (or the reverse) and synthesis works either way.
You select the language with the language request parameter; when you omit it, the model is chosen from the voice’s own locale. simba-3.0 routes English to its English training and everything else to its multilingual one, so this is an internal detail: the model ID you pass never changes.
Languages outside this list often still produce usable audio on simba-3.0, but they are not validated or officially supported. For any other language, use simba-3.5.
Fully Supported Languages
The following table describes the legacy simba-multilingual model, not Simba 3. It requires an API version pinned before 2026-09-21; from 2026-11-21, that ID is served by Simba 3.5. For current integrations, use the Simba 3.5 languages above.
Beta Languages
These are the beta languages of the legacy simba-multilingual model. They are not a list of supported Simba 3 languages. The same retirement (2026-09-21) and replacement-training (2026-11-21) dates apply.
Using the language Parameter
Our speech synthesis endpoints (/v1/audio/speech and
/v1/audio/stream) support the optional language
parameter, which should follow the locale naming standard (e.g., en-US,
fr-FR).
When to specify the language:
- Known single language: If you know the input text is entirely in one language, providing the
languageparameter will result in better audio quality. - Unknown or mixed language: If you’re unsure of the input language or the text contains multiple languages, omit the
languageparameter. Speechify models will automatically detect and handle the language(s) in the input.
Voice Cloning Languages
All three Simba 3 models support voice cloning on paid plans. A cloned voice can synthesize in the selected model’s supported languages: the 48 above on Simba 3.5, English on Simba 3.2, or the six above on Simba 3.0. Check voice and model compatibility before sending a request. See Voice Cloning for sample requirements and implementation details.