Product

Translate documents into
your mother tongue.

Tarjumi translates documents from English into local African languages. Kikuyu, Swahili, Luo, Luhya, Kalenjin, and more. Google Translate doesn't support most of these. We do.

What you get from this page

A working translator for local African languages. Paste your text, select the target language, and get a translation. Also works as document input — translate contracts, health guides, educational materials, or government notices into the language your community actually speaks.

Translate
Text, article, or document. Any language.

Paste text, drop in a news article URL, or type directly. Select your target language. Tarjumi handles the rest.

Source: English
Translate to:
Translation will appear here...
How Tarjumi Works
Understanding first. Translation second.

Most translation tools do word substitution. Tarjumi understands the meaning of your text across languages using cross-lingual embeddings, then generates a natural translation in the target language. The difference: it knows that "dawa" and "medicine" mean the same thing without looking up a dictionary.

01 — Detect
Identify the language

Automatic language detection for 75+ languages, including Kikuyu and other Bantu languages that standard detectors miss entirely.

02 — Understand
Cross-lingual meaning

LaBSE embeddings map your text into a universal meaning space shared across 109 languages. Kikuyu and English live in the same vector space.

03 — Translate
Generate natural output

Neural machine translation produces fluent output in your target language. Not word-for-word substitution — actual sentences that sound natural.

04 — Verify
Convergent quality check

Back-translation and embedding similarity scoring verify the translation preserved meaning. If quality is uncertain, Tarjumi tells you.

Languages
The languages Google forgot.

42 million Kenyans speak a mother tongue that isn't fully supported by major translation platforms. Tarjumi starts with Kenyan languages and expands across the continent.

Kikuyu
Gĩkũyũ
8.1M speakers
Swahili
Kiswahili
100M+ speakers
Luo
Dholuo
4.4M speakers
Luhya
Oluluhya
6.8M speakers
Kalenjin
Kalenjin
5.4M speakers
Kamba
Kikamba
4.7M speakers
Meru
Kĩmĩĩrũ
2.1M speakers
Maasai
Maa
1.2M speakers
Somali
Soomaali
22M speakers
Amharic
አማርኛ
32M speakers
Yoruba
Ẹdè Yorùbá
47M speakers
Zulu
isiZulu
12M speakers
Use Cases
Knowledge shouldn't require English.
Health guides

Translate maternal health guides, vaccination schedules, and disease prevention materials into the language community health workers actually use with patients.

Education

Textbooks, exam materials, and government curricula translated for students learning in their mother tongue. Comprehension improves when the language matches the student.

Legal documents

Contracts, land titles, and government notices in the language of the people they affect. Understanding your rights shouldn't require a translator.

Emergency alerts

Flood warnings, disease outbreaks, evacuation notices. When minutes matter, people respond faster in their first language.

Agricultural extension

Farming guides, weather advisories, market prices — translated for the farmers who need them most. A Kikuyu farmer reads Kikuyu, not English.

Government services

Huduma Centre forms, eCitizen instructions, KRA guidelines. Public services should be accessible in every official and national language.

Built for the languages that built us.

The word "tarjumi" comes from Arabic (ترجمة) via Swahili — meaning "translation." It's fitting: a word that has already crossed languages now names a tool that helps others do the same. Every language carries knowledge, culture, and identity. When we translate, we don't just convert words — we open doors.

Tech Stack
Cross-lingual AI. No dictionary lookups.

Tarjumi uses neural models that understand meaning across languages — not lookup tables. The same embedding space holds Kikuyu and English, so the system knows they're related before translating a single word.

Understanding

LaBSE

Language-agnostic BERT Sentence Embeddings. 109 languages in one vector space. Swahili trains the model; Kikuyu transfers via Bantu family similarity.

Translation

NLLB-200

Meta's No Language Left Behind model. 200 languages including Kikuyu (kik), Swahili (swh), Luo (luo), Somali (som), Amharic (amh), and more.

Detection

Lingua + Heuristics

Automatic language identification with custom heuristics for Bantu languages. Detects Kikuyu via diacritical markers that standard detectors miss.

Verification

Convergent QA

Back-translation + embedding similarity as dual-path quality check. If two independent paths agree the meaning was preserved, the translation holds.

Your language is not a barrier
to knowledge.

Tarjumi is built by CIF AI. We're starting with Kenyan languages and expanding across the continent. If your language isn't listed yet, reach out — we'll add it.