Voice
Record short clips in your language.
Read a prompt aloud, tell a story about your grandmother's cooking, or narrate a short scene. Payouts are highest here because native voice data is the hardest to buy.
Made in Canada · For Africa
Get paid to share African culture.
You already speak the language. We turn that into income.
Early access — join the waitlist.
Paid direct to your bank · Paystack · Live in Nigeria today
Early access · Live stats · Updated AUGUST 2026
₦0Paid
0Tasks completed
0Active earners
0Countries
Numbers refresh in real time from Paystack payout records.
Train the AI of tomorrow
Silicon Valley trained AI on the wrong internet.
The chatbots you use every day were trained mostly in English, mostly in the US, mostly by people who've never had to explain what jollof means. AfriTask AI pays Africans to fix that — one voice clip, one translation, one honest label at a time. Africa teaches the model. The model pays Africa back.
Your language · Your culture
Six kinds of knowledge Silicon Valley can't buy.
Not survey data. Not stock photos. The stuff that lives in kitchens, buses, and voice notes.
- 01
Voice dialects
Yoruba, Igbo, Hausa, Pidgin, Swahili, Amharic — spoken naturally, not read from a script.
- 02
Cultural context
Proverbs, folklore, history, wedding rituals, market bargaining. The knowledge search engines can't scrape.
- 03
Food knowledge
Jollof, egusi, injera, suya — real names, regional preparations, and the arguments behind them.
- 04
Everyday translations
How people actually text, joke, and argue across 8+ African languages. Formal and street.
- 05
Modern slang
Naija Twitter, Kenyan lingo, Ghanaian pidgin. Six months out of date and the model already sounds foreign.
- 06
Visual context
Markets, danfos, aso ebi, danfo drivers, and daily life. Framed by locals, not stock.
Task types
Different tasks, different weights, different pay.
Translation
Translate short lines both ways.
Labeling
Tag photos with what you actually see.
Rating
Judge whether an AI's answer sounds African.
Chatbots trained on Silicon Valley text get the tone wrong. You catch it in one read.
How you get paid
Three steps.
No middlemen.
- 01
Enter a round.
Every task is a small round with a fixed reward pool. You spend a few Work Points to enter — that keeps quality high and locks out bots.
- 02
Do the task honestly.
Follow the instructions on the screen. Real work in real time. No copy-paste from ChatGPT — the grader detects it and disqualifies you.
- 03
Get paid.
Top scorers are ranked by our grader. Payouts land in your bank via Paystack — clearing same-day in Nigeria today, with more African markets coming.
For enterprises
Building AI for African users? Start with real African data.
Voice, language, and cultural datasets — collected ethically, license-clean, delivered to spec. Working with African labs, product teams, and researchers who need training data that isn't a translation of English.
enterprise@afritaskai.caVoice datasets
Naturally-spoken African-language corpora. Quality-graded, license-clean, delivered as WAV + transcript.
Language datasets
Parallel corpora for machine translation, TTS, and transcription across 8+ African tongues.
Cultural context
Structured cultural knowledge — food, tradition, slang, folklore — as JSONL for RLHF and eval sets.
FAQ
Straight answers.
Ready when you are
Your first payout could clear tonight.
Sign up with your phone number. First task is on the house.