Mozilla Common Voice

An open voice dataset, so that speech recognition works in every language.

A commons for AI Solidarity and inclusion Digital commons

Volunteers record and check sentences in their own language. The corpus is released into the public domain, including for languages the big platforms ignore.

And where does AI come in?

The project does not build AI itself: it produces the common good that AI needs. Without open data, speech recognition stays in the hands of a few companies and a few languages.

The AI label

Model or technique: No model of its own: the corpus is used to train other people's speech recognition.
Model published: Not filled in
Where it runs: No inference: the site collects and checks recordings.
Data: Sentences read by volunteers in nearly 300 languages; corpus published under CC0.
Role of people: Other volunteers listen to every recording, which is validated after at least two approvals.
Frugality: Not filled in

“Not filled in”: MAUB could not check it with the project. Know more? Write to us.

Something wrong or out of date? Write to contact@subalpine.it.