Under each word on the dials you'll see a phonetic spelling like /li.βeɾˈtað/. That's the International Phonetic Alphabet — one symbol for one sound, so you always know how a word is said, not just how it's spelled. Most symbols are the letters you already know. This page is the handful that aren't. Every line has two ▶ buttons: sound says the bare sound on its own, word says the whole example word.
These symbols aren't just for reading — they're the checklist Elena's new ear listens with. Speech recognition is forgiving by design: say a flat English d in todo and it still hears todo. So alongside it we're building a second listener that keeps your raw audio and asks how each sound actually came out.
Because Elena set the sentence, every sound you meant to make is known in advance. A neural acoustic model pins each one to its instant in your recording and scores it against the native target — paying special attention to exactly the sounds on this page: one tap where perro needs the trill, a hard d where todo wants ð, vowels that glide instead of staying pure. Your melody and rhythm are compared against Elena's own rendering of the same sentence.
Read the word, say it in your head, then tap the blurred spelling to check the sounds.