Voice data collection · English · Yorùbá · Hausa · Igbo

Build speech tools with local voices.

Contributors submit consented recordings for independent human review. Approval creates an eligible dataset record; model training and WAXAL/FLEURS workflows are separate steps.

Storage and privacy: This static page does not store, transmit, approve, or export recordings by itself. Configure a secured API before using it for real submissions. The API must enforce identity, consent, independent review, rights, retention, and access controls.

Submit a recording

Record in your browser, listen back, then submit only when your API is configured.

Assigned contributor code Language Recording (1 second to 2 minutes)

Transcript The speaker gave informed consent for this recording to be reviewed for speech research. I have the right to submit this recording and transcript for the stated use.

Your recording remains in this browser until the secured API is configured and you submit. Playback is local.

Independent reviewer portal

Reviewers need a separate authenticated identity. Static HTML cannot protect reviewer access; sign-in and permissions must be enforced by the API.

Listen, verify transcript, then record an independent decision.

Sample	Language / status	Audio & transcript	Decision

Dataset versionLanguage

I confirmed rights for this intended training use.

After review

Reviewer approves samples and verifies the exact training rights.
Export a language-specific dataset ZIP through the secured service.
Run the existing ASR training recipe in an approved GPU environment; FLEURS stays held out for evaluation.
Evaluate the checkpoint before separately approving deployment to an AI product.

WAXAL is not imported automatically. Its specific data and derivative terms need review before use.

Cassava post-harvest guidance for Umuahia

Build speech tools with local voices.

Submit a recording

Independent reviewer portal

After review