Practice on SayMei is built on one object: a dialogue. A dialogue is a short conversation between two speakers, graded to an HSK level, with audio and with characters, tone-marked pinyin, and English lined up underneath. The same dialogue can be listened to, read, or spoken — which mode you pick should depend on what is currently going wrong.
Practice and lessons answer different questions. A dialogue is rehearsal material: nobody teaches you anything, you simply meet Chinese repeatedly and find out what you can and cannot handle. A lesson in the lesson library is instruction: it names a pattern, drills it, and makes you use it. Rehearsal without instruction leaves you fluent in your own mistakes; instruction without rehearsal leaves you able to explain a rule you cannot say. Most learners need both in the same week, which is why the two catalogs sit side by side.
Listening, reading, and reading your lines out loud are free and unlimited, and listening needs no account at all. A free profile adds saved vocabulary and progress. Only live unscripted conversation is metered, because it runs a voice model in real time.
The three modes work best in sequence on the same dialogue rather than separately. Listen once cold. Read the transcript to find the words you missed, which is more often a word boundary than an unknown word. Speak the lines back. Save the new vocabulary to flashcards so it comes round again on schedule. Then, when a grammar pattern keeps tripping you rather than a word, take it into the grammar lessons, where each lesson isolates one pattern.
Listening practice plays a dialogue and lets you hide or show characters, pinyin, and English while you work out what was said. Speaking practice hands you one speaker's role and checks what you say against the pinyin on screen. Both use the same dialogues, so you can run one straight after the other.
Listening and reading dialogues need no account at all. A free profile adds saved vocabulary, progress tracking, and flashcards. Only unscripted live conversation with Mei Lin is metered, because it runs a real-time voice model rather than a stored recording.
A short conversation between two speakers, graded to an HSK level, with generated audio and three aligned text layers: Chinese characters, tone-marked pinyin, and an English translation. Each layer can be hidden, which is what lets one dialogue serve listening, reading, and speaking practice.
One dialogue worked through properly beats several skimmed. Listening once cold, reading the transcript, and then speaking the lines back usually takes ten to fifteen minutes, and that cycle is more useful than replaying audio passively for an hour.