Norynta Data › Language data index › Abaza
Abaza speech data for AI
How much open Abaza speech data exists, where the gaps are, and how to get consented Abaza data collected.
That is very little: under 10 hours is not enough to train or fairly evaluate a speech model. Abaza is one of 223 of the 445 Common Voice languages below 10 validated hours.
Source: Mozilla Common Voice language statistics, fetched 2026-11-05. Figures cover the open Common Voice corpus only.
Need Abaza data?
We recruit native Abaza speakers and collect to your spec: read prompts or natural two-person conversation, transcripts in the native script, speaker age, gender and region labels, noise and synthetic-voice checks, and a consent record for every file. Shared or exclusive licence.
Speak Abaza?
Abaza
Join the contributor waitlist and we'll email you paid Abaza recording and writing projects when they open. You record on your phone and are paid once your work is accepted.
Questions
Where can I get Abaza speech data for AI training?
Norynta collects Abaza speech data to order from consenting native speakers: read prompts, natural conversations, transcripts and private evaluation sets, with a consent and origin record for every file. Request it at collect.norynta.com/buy. For open data, Mozilla Common Voice has 3.6 validated hours of Abaza (as of 2026-11-05).
How much open Abaza speech data is available?
As of 2026-11-05, Mozilla Common Voice lists 3.6 validated hours and 24 recorded hours of Abaza from 65 speakers, ranking 241 of 445 languages by validated hours.
Can I get paid to record Abaza?
Yes. Native Abaza speakers can join Norynta's contributor waitlist at collect.norynta.com/join. We email paid recording and writing projects in Abaza when they open, and pay once the work is accepted.