What are we doing here?
You're helping teach a computer to understand Papiamentu. You read a sentence aloud, the computer tries to write it down, and you correct it. Every recording you Save becomes training data that makes the model better — so the more you do, the more useful it is.
The steps
-
Allow the microphone the first time the browser asks.Press “Allow”. The app can't record without it.
-
Read the sentence shown at the top, out loud. 📖Say it clearly and naturally, the way you'd normally speak.
-
Press ● Graba (Record), speak, then press ■ Stòp (Stop).You can play the audio back to check it before continuing.
-
Press Transkribí (Transcribe).The computer writes down what it heard. The very first one is slow (it loads the model) — just wait a moment.
-
Check the Correct transcription box. Only fix it if you misspoke.It's already filled in with the sentence — change it only if what you actually said was different.
-
Press 💾 Save to dataset, then repeat.It saves your recording and jumps to the next sentence automatically. ¡Danki!
Good to know
🤫 Find a quiet spot. Record with as little background noise as possible.
🗣️ Speak naturally. Not too fast, not too slow — your normal voice.
🔁 More sentences = a better model. Every clip you save helps. Keep going with Siguiente › (Next).
✎ Free speech mode. To say your own words instead of a prompt, press ✎ Papia libre (Free speech), record, then type the correct text yourself before saving.
The score colours (optional)
Press Kalkulá score (WER) to see how well the computer did against your correct text. You don't have to — it's just feedback.
Green — correct word.
Yellow — wrong word (substitution).
Red — the computer missed a word.
Blue — an extra word the computer added.