THE SHORT ANSWER
Roblox has enabled automatic translation for its Text-to-Speech API, supporting 11 languages. The AutoLocalize property is on by default, translating dynamic TTS audio to each player's language setting. TTS assets and six platform languages are not yet supported.
- AutoLocalize property enabled by default on AudioTextToSpeech instances
- 11 languages supported: Arabic, Chinese (Mandarin), English, French, German, Hindi, Italian, Japanese, Korean, Portuguese, Spanish
- Dynamic audio translation works now; TTS asset translation planned for 2027
- Voice characteristics (e.g., gender) preserved across language routing
- Six Roblox platform languages still lack TTS support including Russian and Dutch
What Changed
Roblox has switched on automatic translation for its Text-to-Speech API as of October 8, 2026. The AudioTextToSpeech instance now carries an AutoLocalize property that defaults to true. When enabled, the engine detects the language of the input text, checks each connected player's language setting, and if they differ — and both languages are in the supported set — it translates the text and routes playback to a voice that matches the original voice's characteristics as closely as possible. A female English voice, for example, will map to a female voice in Spanish or Japanese.
Languages Covered and Gaps
The feature launches with 11 languages: Arabic, Chinese (Mandarin), English, French, German, Hindi, Italian, Japanese, Korean, Portuguese, and Spanish. That leaves six languages the Roblox platform already supports — including Russian, Dutch, Turkish, Polish, Vietnamese, and Indonesian — without TTS coverage. Roblox says it plans to expand TTS language parity with the platform's automatic translation system, but no timeline was given beyond "continue to expand."
Dynamic Audio Only for Now
Automatic translation applies only to dynamic audio generated at runtime via AudioTextToSpeech. Pre-baked TTS audio assets created through the creator tools are not translated yet. The announcement notes support for TTS asset translation and generic audio assets is targeted for 2027. Developers who rely on stored voice lines will need to keep managing localization manually or switch to dynamic generation.
Developer Reaction
Early forum response shows practical interest. The team behind Clip It, a game that makes heavy use of TTS, called the update "incredible" for regional reach. Others surfaced existing pain points: TTS still does not run on the server side outside of Studio playtesting unless driven from a LocalScript, and several voices produce a noticeable beep at the end of generated clips. A few creators also flagged the missing languages — Dutch for language-learning scenarios, Russian and other CIS languages for a large player base — as blockers for full adoption.
Implementation Notes
Creators who want the old single-language behavior can set AutoLocalize to false on the AudioTextToSpeech instance. The API documentation and a tutorial have been updated with the new property and translation flow. No migration is required; the default-on behavior means existing dynamic TTS calls will immediately start localizing for players whose language settings differ from the input text.