Our Words
Amina is the community-driven recording platform that transforms native speakers into data scientists. Every recording is validated against the engine's predictions, creating the world's first speaker-calibrated Bantu language corpus.
What is Amina?
Amina (amina.ai) is a Progressive Web App designed for native speakers to record structured language data from their phones — anywhere, anytime.
Unlike crowd-sourcing platforms that collect raw, unstructured audio, Amina guides speakers through a precise recording pipeline where every utterance maps to a specific linguistic prediction from the BTS engine.
The result is not just audio — it is acoustically validated, speaker-attributed, linguistically decomposed training data.
Platform Specs
How It Works
Record Syllables
Speaker records all 140+ isolated syllables for their language. This builds their unique acoustic fingerprint — the calibration baseline.
Build Speaker Profile
System computes F₀ baselines, onset perturbation factors, and duration constants for this specific speaker's voice.
Record Sentences
Speaker records verb forms and sentences generated by the BTS engine, each carrying embedded tonal predictions.
Validate
Automated pipeline compares acoustic measurements against engine predictions. Pass → certified. Fail → re-prompted with corrective guidance.
Calibrate
Validated recordings feed back into the system, refining speaker profiles and strengthening the acoustic ground truth.
The Speaker Community
Gamified Contribution
Speaker Profiles
Syllable recordings calibrate speaker-specific acoustic constants — transforming textbook estimates into physical measurements.
Quality Pipeline
Acoustic Validation
SNR check, pitch tracking, duration validation, nasal bridge auditing — all automated against engine predictions.
Peer Review
Native speaker reviewers verify naturalness and correctness. Multiple reviewers per recording for critical data.
Re-Prompt Queue
Failed recordings generate specific corrective instructions. Speakers re-record with targeted guidance until the thesis passes.
For Speakers
Join as a contributor. Record your language, earn rewards, and help build AI that understands your mother tongue.
Join Amina →For Enterprise
Access the validated, speaker-attributed audio corpus. Every recording linked to acoustic profiles and linguistic certification.
See Data Access Plans →