For many Chinese learners, tones feel like a complete mess—not just for beginners, but even for fluent speakers who have studied for years and still get corrected constantly.

If any of the following sounds painfully familiar, this app was built for you.

Fine in isolation, a mess in sentences.
Reading slowly, character by character, you can nail all four tones. The moment you say a three-syllable phrase, they blur together. Every level starts with the single character, so beginners always have solid ground to stand on. Then that same tone comes back inside a natural short phrase, with one key word hidden, and your ear has to pick it. Short frames first; longer sentences only once the short ones feel easy.

Tone struggles are personal, so there are no generic lessons.
Everyone's blind spots differ: some people say mǎi (buy) as mài (sell); some slide downward every time they aim for a 1st tone; some can't hear the difference between certain tones at all.
That's why progress is level-based instead of one "Intro to the 4 Tones" lecture. Each level drills one specific contrast — obvious differences first, then the high-frequency stumbling blocks, then tone softening in fast connected speech. It covers what most courses skip: 2nd versus 3rd tone, the half-3rd that stops before it rises, the neutral tone at the end of a phrase.

You don't need to self-diagnose.
Your weak spots surface on their own as you play. When a contrast trips you up, we don't expect you to master it in one try — the system queues a series of levels on that exact weakness until you own it.

Trying to order dumplings (shuǐ jiǎo) and asking for sleep (shuì jiào).
That embarrassment is what the training targets. Every pair here is a "wrong tone, different word" pair — mǎi/mài (buy/sell), shuǐ jiǎo/shuì jiào (dumplings/sleep), tǔ/tú (dirt/picture). Syllable charts don't make tones stick. Ordering a plate of sleep does.

Without feedback, you're practicing your mistakes.
Study alone and you repeat a line, decide it sounded fine, and build muscle memory for the wrong pronunciation.
Why no AI scoring? We tested the speech engines on the market and none of them judge tones reliably — say it wrong and you pass, say it right and you fail. So we dropped machine grading. At the end of each level you record yourself, listen back, then immediately hear a native speaker say the same line. Heard back to back, your own ears catch the gap.

You memorize that jiē is 1st tone, and lose it three days later.
Tones learned as isolated trivia rarely survive a week. Here the same words return in different contexts across levels, so the tone binds to the word's meaning and use. You remember the word's melody, not a chart.

Tone sandhi makes you freeze mid-sentence.
The half-3rd tone, consecutive 3rd tones, the way tones weaken in real speech — run that arithmetic while talking and fluency crawls. These shouldn't be memorized mechanically, so there is dedicated short-phrase audio practice that builds the intuition by ear instead of by rule.

No frustration, no burnout.
The design is micro-steps, reinforced often. No theory lectures, no mā má mǎ mà, no eight questions on the same word. A quick warm-up, six listening rounds, four speaking takes, and the day is done. Hard contrasts return later in different formats, so they lock in without feeling like revision.

Every voice you hear is a real person.
No text-to-speech, no robotic AI voices, no "Frankenstein words" stitched from isolated syllables. If the reference audio isn't a human speaking naturally, your ear is aiming at the wrong target from day one.

Levels 1 to 10 are free, with no account needed. Premium unlocks level 11 onward, as a monthly subscription or a one-time lifetime purchase. Every word includes an English translation, and the microphone is used only to record your own practice.

If you've worked on your tones for a long time and still struggle, this is the missing piece.