Simba is the family of text-to-speech models Speechify builds and trains itself. Simba 3.5, launched on 21 October 2026, reads 48 languages under a single model, up from six on Simba 3.0, so the same voice and the same request work whether the text is in English, Japanese or Georgian. Developers can use it today through the SpeechifyAI text-to-speech API.
What Simba 3.5 does
- 48 languages, one model. Regional variants are included, so Brazilian and European Portuguese are both covered.
- Any voice, any language. Every voice in the catalog can read any of the 48 languages, and so can a voice cloned on a paid SpeechifyAI plan. A voice keeps the accent of the recording it was made from, so for a native accent, pick a voice recorded in that language.
- Same voices, same setup. Voice IDs, audio formats, word timings and the SSML that Simba 3.2 applies carry over unchanged.
How it sounds
In blind listening tests in the six languages both models speak, native listeners preferred Simba 3.5 to Simba 3.0 Multilingual in 62% of comparisons (October 2026). The method and the numbers are in the Simba 3.5 fact sheet (PDF).
The quickest way to judge it is to listen. The Simba 3.5 launch page has a short film in which none of the voices are human, and a sample in each of the 48 languages, each read by a voice native to it.
The 48 languages
Arabic, Belarusian, Bengali, Bosnian, Burmese, Catalan, Chinese, Croatian, Czech, Danish, Dutch, English, Finnish, French, Georgian, German, Greek, Gujarati, Hebrew, Hindi, Hungarian, Indonesian, Italian, Japanese, Kazakh, Korean, Latin, Malay, Marathi, Norwegian, Persian, Polish, Portuguese, Romanian, Russian, Serbian, Slovak, Slovenian, Spanish, Swedish, Tagalog, Tamil, Telugu, Thai, Turkish, Ukrainian, Urdu and Vietnamese.
For developers
Simba 3.5 is available on every SpeechifyAI plan, including Free, as the model id simba-3.5 on the speech, streaming and WebSocket endpoints.
- Set model to simba-3.5 and language to the language of your text, for example ja-JP.
- Simba 3.2 and Simba 3.0 stay available, so you can move one call at a time. A request that names no model still uses Simba 3.0.
- Until 21 November 2026, Simba 3.5 bills at the Simba 3.2 rates. Current prices are on the SpeechifyAI pricing page.
The full walkthrough, with a request you can copy, is in the Simba 3.5 announcement on the SpeechifyAI blog, and the reference is the models guide in the SpeechifyAI docs.
Frequently asked questions
Simba 3.5 is Speechify's text-to-speech model for 48 languages, launched on 21 October 2026. Developers call it as simba-3.5 on the SpeechifyAI API.
48, from Arabic to Vietnamese, including regional variants such as Brazilian and European Portuguese. Any voice in the catalog can read any of them.
Yes. A voice cloned on a paid SpeechifyAI plan works on Simba 3.5 with the same voice ID, in any of its 48 languages.
Listen on the Simba 3.5 launch page at speechify.ai, or create a free SpeechifyAI account and call the API with model set to simba-3.5.
