Mac Live Caption Languages: Recognition vs Translation
See why spoken-language recognition and translation targets differ, with Alison Workspace’s 101 recognition choices and up to 21 translation targets.
2026-08-31 · 12 minThe short answer
A Mac caption app does not have one universal language list. It first recognizes the language being spoken, then optionally translates the recognized text into a target language. Those are separate capabilities with different catalogs. As of August 31, 2026, Alison Workspace offers 101 recognition choices covering 100 spoken languages and up to 21 translation targets. Actual translation availability is checked on the user’s Mac and can vary with macOS version, region, and installed Apple language resources.
Recognition language and translation target are different settings
The spoken-language setting tells the speech engine what it should listen for. The translation setting tells the app which language you want to read. A language may be available for recognition but not as a translation target, so captions can still appear in the original language even when a translated line is unavailable. Automatic detection can be useful when the source is unknown, but selecting the known language generally gives the recognizer a clearer starting point. A supported label also does not guarantee identical accuracy across accents, speakers, or recording conditions.
The 101 recognition choices
The current recognition catalog is: 简体中文, 繁體中文, 廣東話, English, 日本語, 한국어, Français, Deutsch, Italiano, Português, Español, Русский, Türkçe, Polski, Català, Nederlands, العربية, Svenska, Bahasa Indonesia, हिन्दी, Suomi, Tiếng Việt, עברית, Українська, Ελληνικά, Bahasa Melayu, Čeština, Română, Dansk, Magyar, தமிழ், Norsk, ไทย, اردو, Hrvatski, Български, Lietuvių, Latina, Māori, മലയാളം, Cymraeg, Slovenčina, తెలుగు, فارسی, Latviešu, বাংলা, Српски, Azərbaycanca, Slovenščina, ಕನ್ನಡ, Eesti, Македонски, Brezhoneg, Euskara, Íslenska, Հայերեն, नेपाली, Монгол, Bosanski, Қазақша, Shqip, Kiswahili, Galego, मराठी, ਪੰਜਾਬੀ, සිංහල, ខ្មែរ, ChiShona, Yorùbá, Soomaali, Afrikaans, Occitan, ქართული, Беларуская, Тоҷикӣ, سنڌي, ગુજરાતી, አማርኛ, ייִדיש, ລາວ, O‘zbek, Føroyskt, Kreyòl Ayisyen, پښتو, Türkmençe, Nynorsk, Malti, संस्कृतम्, Lëtzebuergesch, မြန်မာ, བོད་སྐད་, Tagalog, Malagasy, অসমীয়া, Татарча, ʻŌlelo Hawaiʻi, Lingála, Hausa, Башҡортса, Basa Jawa, and Basa Sunda. The list covers languages and some distinct written or spoken varieties; support means the engine can be selected, not that every variety has the same benchmark score.
The 21 translation targets
The current translation target choices are: 简体中文, 繁體中文, English, English (UK), 日本語, 한국어, العربية, Deutsch, Español, Français, हिन्दी, Bahasa Indonesia, Italiano, Nederlands, Polski, Português, Русский, ไทย, Türkçe, Українська, and Tiếng Việt. Recognition and translation are configured separately. A source language outside this target list may still be recognized and shown as original captions. Translation resources are checked dynamically on the Mac, so a listed target can still require a download or be unavailable on a particular macOS version or region.
Why Apple Speech and local Whisper can show different coverage
Alison Workspace can use more than one recognition route. On macOS 26, an explicitly selected language with an available Apple on-device speech resource can use Apple Speech’s native streaming path. Older systems, automatic detection, languages not available through that Apple route, or missing resources can use local WhisperKit. This broadens recognition coverage, but the engines do not necessarily segment speech, finalize text, or handle every accent in the same way. Translation remains a separate Apple capability checked on the individual Mac.
Offline use and a reliable language-pair test
The first use of a language may need Internet access to download an Apple resource or local Whisper model. After the required resources are installed, audio capture, recognition, and translation run on the Mac. Before a long meeting or class, play one minute of permitted audio with a clear speaker, choose the known source language and desired target, and check a name, number, and technical term in both lines. Repeat with your real meeting app or video player. Code-switching, overlapping voices, music, dialects, and unusual terminology can still reduce accuracy.
Frequently Asked Questions
Why are there more recognition languages than translation languages?
Speech recognition and text translation use different models and resource catalogs. Alison can show original captions for many languages even when that language is not available as a translation target.
Can every recognized language be translated into every target?
Not necessarily. Availability is checked on the Mac and can vary by macOS version, region, installed resources, and the language pair.
Do Mac live captions work offline?
Yes after the required Apple language resource or local Whisper model has been downloaded. The first resource installation may need Internet access.
Should I use automatic detection or select the language?
Select the known spoken language when possible. Automatic detection is useful for unknown input, but a manual choice gives the recognizer a more specific starting point.
Does supported mean perfectly accurate?
No. Accuracy still changes with accent, audio quality, background sound, overlapping speech, names, numbers, and specialist vocabulary.