How to practise speaking a language alone (three modes that work)

"Practise speaking" is treated as one activity, which is why people who have no conversation partner conclude they cannot do it at all. In fact there are three separate exercises inside that phrase, they train different things, and two of them work perfectly well alone.

Mode one: read it aloud

The sentence is in front of you. You say it. That is the whole exercise.

What this trains is the mouth, not the brain — the physical business of producing sounds your language does not contain, in sequences it never asks for. Japanese ryo, Korean tense consonants, the Spanish rolled r, English th. Your tongue has no muscle memory for these and gets it only through repetition.

The reason it is worth doing deliberately rather than assuming it happens by itself: silent reading builds a private, wrong pronunciation that becomes very hard to remove later. If you have read a word two hundred times and said it zero times, you have two hundred repetitions of an error.

This is Amiko's Read along mode: the sentence is shown, you tap and read it aloud, and it checks what you actually produced against what was on screen. The check matters — reading aloud without feedback lets you rehearse a mistake confidently.

Mode two: say it from meaning

You are given the meaning in your own language. You produce the sentence in the target language, out loud, with nothing to copy.

This is a large step up, and it is the one most learners skip. There is no text to lean on, so you have to retrieve the words, choose the grammar, and assemble the whole thing under time pressure — the exact operation that fails you in a real conversation.

If reading aloud trains the mouth, this trains retrieval. And retrieval is what breaks down when someone actually speaks to you: the words are in there, but you have never practised pulling them out at speed.

Amiko calls this Say it. It is noticeably harder than Read along and worth doing on the same material — read the sentence aloud first, then a few minutes later produce it from the meaning alone.

Mode three: free conversation

Someone says something unpredictable and you respond. Nothing is given.

This trains the thing the other two cannot: handling what you did not plan for. Real conversation constantly asks you to understand a question you were not expecting and answer it before the pause gets awkward. No amount of scripted practice prepares you for that, because the whole difficulty is the absence of a script.

This is the mode that traditionally requires a partner, and the one people therefore never do. An AI roleplay conversation fills the gap: it answers unpredictably, it stays in character, and you can be wrong forty times in a row without spending anyone's patience.

Use all three, in that order

The three build on each other, and the order is the useful part.

Take a sentence pattern you want. Read it aloud until your mouth stops fumbling. Then produce it from meaning, until it comes without hesitation. Then go into a conversation and try to use it, where it will come out slightly wrong because the situation is different — which is exactly the point, and where the learning finishes.

Skipping to conversation means fumbling the sounds while also under pressure. Never leaving mode one means a beautiful accent and total paralysis.

How to check yourself without a teacher

The honest limitation of solo practice is that you cannot hear your own errors — your brain plays back what you meant to say, not what you said.

Two things get around it. Record yourself and listen a day later, when the intention has faded and you hear the actual sound. Or use automatic pronunciation checking, which is unsentimental about what you produced.

Neither replaces a good teacher's ear. Both are enormously better than the alternative, which is fifty repetitions of the same mistake.

Five minutes, three modes

Two minutes reading aloud, two minutes producing from meaning, one conversation. Daily.

It is a small routine and it covers all three skills, which is more than most people manage in an hour, because most people spend the hour doing exercises where they never open their mouth.