Speech AI doesn't understand how Africa speaks French.
We measure the gap, and we build the data that closes it.
The problem
Over 100 million people in Africa speak French every day, with their own accents and local words mixed in.
Most speech AI is trained on French from France.
Almost no public data exists for African French, so the errors stay hidden until your customers hit them.
What we do
Evaluation.
Test your speech model on real African French and see exactly where it fails: accents, numbers, place names, code-switching.
Datasets.
Licensed speech data with exact transcripts and speaker information.
Custom collection.
Recordings for your use case, in the cities and languages you need.
How we collect
Local coordinators record real people in their own cities. Every speaker gives consent and is paid. Every transcript is written by a native speaker, word for word, including local words.
Where we are
Pilot underway in Benin.
First benchmark results: end of October 2026.
Contact
Get a free evaluation of your model.
Or write to contact@oralilabs.com