Reflex loop · Shadowing loop
Hear the model → shadow it → get scored
Pick a syllable in the left column, hear the standard and the slow take, shadow it, then read it back to be scored. You get a score, what the machine heard, and replay buttons — for the native model and for your own voice. Free mode uses your browser's recognition and is unlimited; AI-beta gives detailed pronunciation scoring.
Practice bench
Choose the sound you want to work on today
One sound per round. Read slowly until your mouth finds the shape on its own, then speed up — and always replay the slow model after a score.
Pick a sound to practise
Six common sounds from each stage, or type the exact one that keeps tripping you up.
Hear the standard
Load the model at natural speed before reading along. Do not guess the sound — hear it first.
Inspect the slow take
The slow version exposes the puff of air, the vocal-fold buzz and the pitch glide.
Shadow it
Speak along while the audio is still running — do not wait for the end. Your mouth learns faster than your memory.
Read it back for a score
Tap the mic, read aloud, get your score and replay your voice next to the model.
Shadowing tips for beginners
Do not translate in your head. Imitate the music of the syllable: its height, its length, its puff — like mimicking a line of a song. For tone 3, exaggerate the dip (sink really low, then flick up); for tone 4, be decisive, like giving an order. Ten sounds a day and in two weeks your ear separates z/zh/j on its own.