You do not lack the words for 你好 (nǐ hǎo, hello). You lack a place where that line leaves your mouth when nobody is on the other end. This is not the general self-study map. That map is learn Chinese alone. The job here is speaking: shadowing that produces, a 30-second recording on Monday and Friday, and tone repair.
Passing the written HSK does not test speech. The HSK (Hanyu Shuiping Kaoshi) that Wikipedia HSK summarizes is, in the classic six-level system, listening, reading, and writing. The oral is separate: HSKK. You can know 谢谢 (xièxie, thank you) on a flashcard and go quiet at the till.
The study clock still lives in the 20-minute routine. This page is only the mouth.
Key takeaways
- The block is not ‘I don’t know enough.’ It is not making sound. 你好 and 谢谢 are already enough to start speaking.
- HelloTalk-style exchange fails in month one because it is irregular: time zones, English in return, and no repeat of the same line.
- Useful shadowing is not karaoke. One phrase, copy tone and rhythm, then 30 seconds recorded on Monday and Friday.
- A tutor that remembers your tone 2 vs 3 and 买/卖 pays off when Friday still sounds the same. AI can be wrong; Premium covers real model cost.
In this article
The block is not “I don’t know enough”
The block is fear of sounding wrong, not a hole of 150 words. Classic HSK 1, as chinesetest.cn describes it, is about that many. You do not need them to order at a counter. You need five sentences said until they come out without translating.
Those five, with pinyin: 你好 (nǐ hǎo, hello). 谢谢 (xièxie, thank you). 我要这个 (wǒ yào zhège, I want this). 多少钱 (duōshao qián, how much is it?). And the pair 买 / 卖 (mǎi / mài, to buy / to sell), which is where tone gives you away.
Waiting until you “know more” stretches the silence. Each week of cards-only teaches you to recognize 谢谢 and not to say it. Your brain stores the gloss. Your mouth has no file.
You do not need a partner in month one. You need a script and an ear. The HSK 1 guide says what the paper contains. It will not listen. You can listen to yourself, if you record.
If you can read 我要这个 and cannot say it in three seconds, the hole is output. You are ready for those syllables. You are not ready for a dinner party.
Why language exchange fails in month one
HelloTalk and apps like it are not the villain. Month one’s calendar is. You do not need a user-count: the other person lives in another time zone, wants your language, and you have ten sentences. That chat does not hold.
Exchange is a barter. Half Chinese, half English or Spanish. You leave having explained your job in English and having said 你好 once. That is not output. It is networking.
There is also no repeat. Yesterday they fixed 多少钱. Today a stranger talks about their dog. The tone 3 on 买 never comes back. A tutor with memory puts it in front of you again. An exchange chat does not.
Correction is random. They fix a word you did not ask about. They let a flat tone 2 pass because “it’s understandable.” Month one does not need understandable. It needs 卖 not to sound like 买.
“Sunday?” and Sunday never happens. Two calendars, zero contract. Use exchange from month three, once you have a script. The phrases hub is for picking that script, not for waiting on a chat at 11 p.m.
Shadowing that produces (not karaoke)
Shadowing is speaking over the audio, almost in time, copying rhythm and tone. Karaoke is singing over a three-minute clip and feeling like you “practiced.” Karaoke never isolates 谢谢. Useful shadowing does.
Pick one utterance. Not a vlog. An HSK 1 greeting is enough: 你好，谢谢. Listen twice with your mouth shut. The third time, speak over it. The fourth, the audio stops and you say it alone. That is the work: no net.
Copy three things, not “a Chinese accent.” One: how long each syllable lasts. Two: whether the tone rises, falls, or stays low (tone 3). Three: the phrase stress, not the stress of your English. 我要这个 is not “wo yao this-guh” on a London beat.
If the clip runs longer than eight seconds, cut it. A whole dialogue in month one is karaoke in disguise. You repeat the easy bits (你好) and mumble the hard ones (多少钱). The cut forces the hard line.
After shadowing, you record. No backing track. If you only shadow, your ear leans on the model and never hears your flat tone 3. A phone memo is enough. Native HSK 1 audio, not a news podcast. CLEC backs the teaching frame; it will not send you the clip.
A 30-second recording protocol
Monday and Friday. The same three sentences. One file. Thirty seconds, not a journal. If Friday still sounds flat, do not open a new list: the hole is tone, not inventory.
Pick the script and do not change it for five days. HSK 1 example: 你好。谢谢。我要这个。 If those already come out, swap the third for 这个多少钱？ (zhège duōshao qián, how much is this?). Do not mix 买 and 卖 on the same Monday until you can hear them.
How to record. Phone upright, about 20 cm from your mouth. No music. Say the three lines twice: slow, then till-speed. If you trip, leave the trip in. An honest file teaches more than a pretty eighth take.
How to listen. Friday, headphones, no pinyin on screen. Ask one question: do I recognize my own 谢谢? If not, the flashcards are lying. You recognize the character. You do not produce. Write one line: “flat tone 3” or “多少钱 swallows the second syllable.” One line. Not an essay.
Name the file with the date: 2026-09-14-monday.m4a. Five different Fridays are the oral progress you can show without an exam. The protocol does not replace HSKK; written-test dates live on chinesetest.cn.
If a Friday sounds better, do not reward yourself with 20 new sentences. Reward yourself with the same three, a little faster. Until 我要这个 has tone, speed only pushes the error along.
What to correct: tone 2 vs 3, 买 vs 卖
Do not correct “the accent.” Correct one pair. The pair that gives beginners away: tone 2 (rises) against tone 3 (dips and stays low), and 买 (mǎi, to buy) against 卖 (mài, to sell). The ear lab is Mandarin tones. This post does not clone that guide. You apply one pair to the recording.
Tone 2: 什么 often rises. Tone 3: 好 in 你好 is not a short tone 2. In 你好, the first 你 (nǐ, tone 3) changes by sandhi before another tone 3: it sounds closer to tone 2, and 好 stays low. If both syllables rise, you are not saying 你好. You are saying something else. That shows up on the file, not in the theory.
买 is tone 3. 卖 is tone 4 (a sharp fall). If on Friday both sound like a flat “mai,” the counter cannot tell whether you are buying or selling. Record only the pair: 我要买这个 (wǒ yào mǎi zhège, I want to buy this) and 这个卖多少钱 (zhège mài duōshao qián, how much is this selling for?). Five times each. The rest of the session can wait.
A typical English-speaker miss: treating tone 3 like a quick stressed fall, and making tone 2 just as short. Tone 3 needs time. If you rush 谢谢 to “sound fluent,” you squash xiè and the second syllable disappears. Slow is the fix. Fast is a disguise.
One correction per file. If Monday’s note was “tone 3,” Friday you only listen for that. An ear does not repair two fossils at once. When 买 / 卖 already sound different, stop. The rest of the tones wait for the lab.
When a tutor (human or AI) pays off
A tutor pays off when Friday still sounds the same and you can no longer hear the error. Until then, the file is cheaper. Irregular exchange is not a tutor: it does not remember that you flattened tone 3 on Tuesday.
A human tutor pays off when you already have sentences. You can say 我要这个 and 多少钱. You need someone to stop you in the moment and make you repeat. Two sessions a month, with Friday’s file on the table, beat an hour of free chat where you hide inside 你好.
An AI tutor pays off if it does three things. Text and voice, not typed chat only. Memory of *your* errors (tone 2 vs 3, 买 vs 卖), not a model that greets you from zero every Tuesday. And an honest limit: it can be wrong. It can mark a flat tone 3 as fine. That is why you check a native clip, not your faith.
A generic chat is not that. It does not remember last Friday. It praises “good pronunciation” because the text was right. The characters for 卖 can be perfect and the audio can still be 买. If the tool cannot hear, it is not tutoring speech.
Paying for a model has a real cost. That is why the voice plan is not the lists plan. If you do not yet have three sentences, do not pay: record first. If two Fridays sound the same, a tutor (human, or AI with memory) is the next spend, not another streak app. The official frame does not change: Wikipedia HSK, CLEC, register on chinesetest.cn. The tutor hears you. It does not examine you.
One concrete week without a partner
Monday. 30 seconds: 你好, 谢谢, 我要这个. Two takes. Shadow the same three, eight minutes, one short clip. Stop. Do not open HelloTalk “to see if anyone is there.”
Tuesday. Same three at the start, no recording. Then 多少钱 in shadowing, cut down to one question. You finish by saying 这个多少钱？ three times. You note whether you swallow qián.
Wednesday. Only the 买 / 卖 pair. Five 我要买这个. Five 这个卖多少钱. If they still sound the same, Thursday adds no new sentences. Repeat the pair.
Thursday. Role-play alone. You are the customer. 你好. 我要这个. 多少钱. 谢谢. No reading. If you stall, you go back to Monday’s file, not a new list. The phrases hub can lend a shop line on Saturday, not today.
Friday. 30 seconds with Monday’s script. Headphones. One line of notes. If it sounds the same, next week is the same three. If 我要这个 already has tone, you may raise speed, not inventory.
Saturday. Two lines from the phrase bank, said once each. Sunday: listen to Monday and Friday. One line of notes. SRS, if you still run it, is the 20-minute routine. This week puts a mouth on it. The self-study map remains learn Chinese alone.
A 7-day trial: a tutor that remembers your tones
If Friday has nothing left to teach you, the next step is not another streak. It is someone (or a model) that remembers you flatten tone 3 on 买. Chinese Trainer has a text-and-voice tutor, with memory of your misses, on Premium. Premium costs more because the models cost money. Answers can be wrong: a tone marked “fine” can still be flat. Check it against your file.
The inventory is in what’s inside the app. One step: start the 7-day trial. If that week it hands you the same tone error twice, it is doing the job. If it only chats, cancel before the trial ends. Official HSK registration stays on chinesetest.cn, not here.
Pinyin is the official romanization of Mandarin. Hanzi are the written Chinese characters. HSK is Hanyu Shuiping Kaoshi. Reviewed by Javier Garcia Floriano. About us. Contact.
- Apply one method today, not a second streak.
- Measure one pair or one mock, not XP.
- Use the single CTA on this page.
Frequently asked questions
Can you practice speaking Chinese if you have nobody to talk to?
Yes. Month one is a short script, not a native friend. Shadow one sentence, record 30 seconds on Monday and Friday, and pick one fix (a tone or 买/卖). A partner comes later, once you have something to repeat.
Does HelloTalk or a language exchange work in the first month?
It works later. In month one the chat is irregular: another time zone, they want English, and you will not hear 多少钱 two days in a row with the same correction. Use exchange once you can hold three sentences without looking at pinyin.
Does passing the written HSK mean I can speak?
No. Classic HSK scores listening, reading, and writing. Speaking is a separate exam (HSKK). You can recognize 谢谢 on a card and freeze at the till. The official paper and dates live on chinesetest.cn.
Can an AI tutor replace a native speaker?
No. A tutor (human or AI) that remembers your tone errors beats a generic chat. AI mishears sometimes. Check against a native clip. Premium exists because the models cost real money.
Helpful internal links
Practice it inside the app
If you want to turn what you just read into real Mandarin practice, jump into Chinese Trainer and review it through flashcards, exercises, listening, and the AI tutor.