Coverage · growing
Every language, on the same three modes.
The corpus is not a frozen file — it grows as native contributors record. Each language is measured on the same three modes, so an empty cell is a visible gap, not a hidden one.
4Languages
1145Consented takes
104638Words read
1504.9Minutes of speech
8Native speakers
| Language | Country | Bantu only | Bantu & English | English (accented) | Narration | Minutes | Accent WER |
|---|---|---|---|---|---|---|---|
| Bemba | Zambia | 378 takes 24.3 min |
363 takes 70.2 min |
111 takes 578.2 min |
5 takes 16.9 min |
689.6 | 6.8% |
| Luganda | Uganda | 60 takes 4.7 min |
60 takes 14.4 min |
112 takes 431.0 min |
coming | 450.1 | 15.1% |
| LUN | Zambia | coming | coming | 55 takes 362.2 min |
coming | 362.2 | 17.5% |
| Lozi | Zambia | coming | coming | coming | 1 takes 2.9 min |
2.9 | — |
How to read this
Bounded frame, rising fill.
Three modes × a known set of stories per language is a measurable denominator. You can see exactly which cells are filled with consented native audio and watch the numbers rise — the opposite of an open-ended web scrape.