Pick a dialogue, choose a speaker to play, and talk it through face to face: your partner says a line, then you say yours. SayMei transcribes what you actually said and lines it up against the line on screen, syllable by syllable. A free account gets one session a day, Premium is unlimited, and it never uses lesson minutes.
What the feedback actually tells you
Each syllable comes back matched or not matched, and the line gets an accuracy percentage (matched syllables over expected ones), so you know when to move on. A syllable counts as matched when the transcript gives the same character, or the same pinyin syllable with its tone mark ignored. So this measures whether you produced the right syllables in the right order, not whether your pitch was right. That limit is deliberate. In Mandarin the tone is part of the word: mā (妈) is mother, má (麻) is hemp, mǎ (马) is a horse, mà (骂) is to scold, and a tone verdict inferred from a transcript would be guesswork dressed up as a score. For pitch itself, the Chinese Tone Checker records you and draws your contour against the target shape.
The error almost everyone makes
Third tone is the usual culprit, because the full dipping-then-rising shape drilled in isolation is not what speakers produce inside a sentence. Before another third tone it is said as a rising second tone, which is why 你好 comes out as ní hǎo. Before a first, second, or fourth tone it stays low and does not rise at all. Reading a whole line aloud, rather than one syllable at a time, is what exposes it: the sandhi only exists once the syllables sit next to each other. The syllable match here will not judge that for you, which is why a sound that keeps slipping belongs in the tone tools below.
Scripted speaking versus live conversation
This is scripted speaking: the words are in front of you, you set the pace, it never uses lesson minutes, and you can repeat a line twenty times without anyone waiting. That is the right tool for building the motor habit. It is not a conversation, because nobody asks you an unexpected question. When you want that, talk with Mei Lin, who replies in real time and corrects you mid-sentence; those sessions run on live minutes, described on the pricing page.
How to get more out of a session
Listen to the whole dialogue first in listening practice so you are copying a model rather than guessing. Shadow the audio once before recording. Then speak your role, and when the same syllable keeps failing, leave the dialogue and take that sound to the tone pair trainer: isolated tones are usually fine, and the errors appear where two syllables meet. Afterwards, swap to the other speaker's role and run the dialogue again.
Frequently asked questions
How does SayMei check my Chinese pronunciation?
Your speech is transcribed and lined up against the line on screen, character by character. A character counts as matched when the transcript gives the same character, or the same pinyin syllable with its tone mark ignored; matched characters light up as you read, and the line scores matched characters over expected ones.
Does speaking practice use up my live lesson minutes?
No. Reading dialogue lines aloud never uses lesson minutes: a free account gets one session a day, and Premium makes it unlimited. Only unscripted live conversation with Mei Lin draws on speaking minutes, so within a session you can repeat a difficult line as many times as you like.
Does speaking practice check my tones?
Not directly. The match runs on syllables and ignores the tone mark, so a line can come back fully matched while a tone was off. That is the honest limit of a transcript, and inventing a tone score on top of one would not help you. Pitch is measured in the free Chinese Tone Checker, which records you and compares your contour with the target shape, and drilled in the tone pair trainer.
Do I need a microphone or an app to practice speaking?
A browser and a microphone are enough: a phone or a laptop both work, and nothing is installed. Grant microphone access when your browser asks, pick a dialogue, choose your role, and start reading. Everything happens on the page.
Is this the same as talking to an AI Chinese tutor?
Not quite. Speaking practice is scripted: the lines are in front of you and you control the pace, which is what builds accurate pronunciation. A live lesson with Mei Lin is unscripted conversation with real-time correction. Most learners use the scripted mode to prepare for the live one.