How do you see pronunciation per phoneme, not only the phrase mark?

How to do it in three steps

Do not start with overallScore. Start with the word below 50, then the phone in 0–39 or 40–59.

1. Type the phrase. Record up to 60 s, 8 MB, mic at 15–30 cm, SNR above 15 dB.
2. Open the bad word. Read the phones. The Phone PCC unit is the aligned phone, not the workbook syllable. /aprender/score-por-fonema explains bands 80–100 through 0–39.
3. Train the phone, not the whole phrase. If take confidence is under 0.5, the phone is void: dirty audio. /aprender/quando-desconfiar-do-score.

What to expect: formats, size and time

Same page caps: 8 MB, 60 s, 16 kHz. No IPA keyboard for you to draw the phone. The engine returns the phone it aligned.

We do not publish a per-phone hit-rate percentage beyond the guide bands. Example native overallScore in the guide is 100; example sentence 87; example confidence 0.95 — all at /aprender/quatro-niveis.

When to use the paid API

When the product shows the phone tape on every lesson, without the tab. API JSON, key at runtime. /products and /pricing.

One tape today: free page. A school app: API. Series /aprender.

The free page exists so you can try the flow with a real file, see the result, and only then decide whether the API belongs in the operations pipe. Nothing is published, sent or charged from here: the click goes to the tool, and the account lives on products and pricing when volume no longer fits in the tab. This screen is the rehearsal; the API is the lot, the key and the account quota.

Frequently asked questions