English doesn't reward effort — it rewards precision. JunaTalk listens to every syllable you say and grades it on two signals: the sounds you produced, and where you placed the stress. The red, amber, and green chips tell you exactly which syllable cost you the score.
Sounds. Did your phonemes match the target accent? An /r/ that sounds like an /l/, a missed /θ/ — all caught at the syllable level.
Stress. English is a stress-timed language. Get the wrong syllable emphasized and even perfect sounds read as a foreign accent. JunaTalk compares your prosody to a native speaker on the same phrase.
We take the better of the two so a single weak signal can't tank the whole syllable. You see both meters because both matter — and because knowing which one slipped tells you what to work on.
After each phrase, every syllable is recolored on a 0–100 scale. Green means a native speaker would hear it as that word. Amber means you were close but drifted. Red means the syllable will misread.
Tap any colored syllable to see its individual Sounds and Stress scores. The breakdown tells you whether you said the right thing wrong (sounds), or the wrong thing right (stress).
Whether it's amber, red, or just a curious green — tap any syllable to open the word popover. You'll see your exact Sounds and Stress scores, hear the reference audio at any speed, and start a focused lesson built around that word in one tap.
English speech recognition built for accented voices — not just American newscasters. Trained to hear you correctly even when you're not quite there yet.
Industry-grade phoneme-level scoring (Azure). Each syllable scored against the reference audio of your target accent.
Custom syllable-emphasis comparison vs the native reference's RMS envelope. Catches stress drift even when sounds are perfect.
Asymmetric exponential moving average — fast up (0.75 weight), slow down (0.35). Reflects real learning, not one-off slumps.
30 days free, then $9.99/month. Cancel anytime.
Get early access