All Voices
Languages
7,000+ languages supported
Status
Live
Origin
Built with language communities
Overview
All Voices is African Languages Lab's community platform for building language datasets. Native and fluent speakers translate text, transcribe audio, record speech, describe images and peer-review each other's work, one contribution at a time, for 7,000+ languages worldwide, with a special focus on African languages such as Yoruba, Igbo, Hausa, Swahili, Amharic, Zulu, Twi, Wolof and Fula.
Most of the world's 7,000+ languages have very little data behind them, so AI cannot understand them well. All Voices changes that by putting the people who speak these languages at the centre of the work. Every contribution is checked, through automatic audio quality checks and peer review, because trusted data is what makes better language technology possible.
Contributors earn points and keep streaks as they go. Organisations can also run their own projects on All Voices: upload source material, choose the task types, and receive a reviewed, exportable dataset. Download the app on iOS or Android, or contribute on the web.
Key features
Translate
Read source text and write a translation in your language.
Transcribe
Listen to audio and write what you hear in text.
Record speech
Read a prompt aloud. The app checks audio quality automatically before you submit.
Describe images
View a photo and speak or write what you see in your language.
Validate
Rate other contributors' work. Peer review is what makes the dataset trustworthy.
Earn points and keep your streak
Every task earns points: 10 for a translation or transcription, 15 for a speech recording, 12 for an image description and 5 for a validation, plus +3 for your first contribution of the day and +25 for a 7-day streak.
Built for organisations
Partner institutions get project management tools, access to the contributor network, a peer review pipeline, role-based team access and datasets exportable as JSON or CSV.
By the numbers
7,000+ languages supported worldwide
5 task types: translate, transcribe, record speech, describe images and validate
Automatic audio quality checks on every recording
Peer review built into every dataset
Up to 15 points per contribution, plus daily and 7-day streak bonuses
Datasets exportable as JSON or CSV for partner organisations
