Nine languages. RTL. AI in every one.
Education is not English-only. ArthurAI™ supports nine languages across the entire platform — UI, AI tutor responses, AI-generated curriculum, text-to-speech, and the educator-attestation surfaces. RTL document direction handles Arabic and Urdu. The international footprint is the proof.
English
English · LTR · code
enDefault. Full coverage.
Arabic
العربية · RTL · code
arRTL document direction. Azure Neural Voice ar-SA-ZariyahNeural for TTS.
Spanish
Español · LTR · code
esLatin American Spanish prioritized; ES-Spain variant available. Azure Neural Voice es-ES-ElviraNeural.
Urdu
اردو · RTL · code
urRTL document direction. Pakistan-facing deployments (FDE, KPSIAJ, AI Future Lab, GIK, Kohat, PIHS, PEC, COTHM, ICS+NEI, Alkhidmat KP, Smart Learnify). Azure Neural Voice ur-PK-AsadNeural.
French
Français · LTR · code
frStandard French. Azure Neural Voice fr-FR-DeniseNeural.
Turkish
Türkçe · LTR · code
trStandard Turkish.
Swahili
Kiswahili · LTR · code
swEast Africa-facing deployments (IUCEA, Daystar, OUK, KEPSA, Craft Silicon, Kadisco). Standard Swahili.
Portuguese
Português · LTR · code
ptBrazilian Portuguese prioritized; Continental Portuguese variant available.
Hausa
Hausa · LTR · code
haWest Africa-facing deployments (TechMindset Africa, SO TECHAFRICA Nigeria).
How language flows through the platform
- UI translation. The unified portal ships with i18next configured with HTTP backend and namespaced JSON translation bundles. Locale detection runs against the user's saved preference, falling back to the institution default and then to English.
- RTL document direction. Arabic and Urdu locales flip the document direction to RTL automatically. Layout mirrors, not just text. Tested across the unified portal and per-edition portals.
- AI tutor responses. The tutor's system prompt includes the learner's language preference as context. The tutor responds in the learner's preferred language directly through multilingual frontier models that handle it natively — no translate-then-respond round-trip.
- AI-generated curriculum. Course generation accepts a `languageToLearnIn` parameter. Curriculum, lesson content, and assessment items generate in the target language with the same 6-step structure.
- Text-to-speech. Azure Cognitive Services Speech provides Azure Neural Voices for English (Jenny), Arabic (Zariyah), Spanish (Elvira), Urdu (Asad), and French (Denise). Per-section ReadAloud, floating audio controls, and language-aware voice selection.
- Translation service. A dedicated translation pipeline with token-tracking and JSON-fidelity validation handles bulk content translation for institutional onboarding when an institution wants English-source curriculum delivered in a non-English target.
- Locale-aware formatting. Dates, numbers, and currencies use the Intl API with locale-aware formatters; date-fns locales are loaded per language.
- Per-edition language tuning. Each edition portal can ship its own translation bundles when terminology differs (e.g., CLE uses Employee / Trainer / Organization, others use Student / Teacher / School).
Why nine, not five hundred
Adding a language to ArthurAI™ is not a checkbox. Each new language requires translation of the UI namespaces, verification of RTL or LTR rendering, validation that the frontier models generate quality content in the target language, integration of the Azure Neural Voice (or alternative TTS), and educator review of the AI-generated samples. The nine we ship are the nine our deployment footprint actually needs. Adding additional languages is on the roadmap as new institutional partnerships justify the work.