Record a 90-second voice note today explaining something you know well, like your commute. This baseline shows your real pauses and restarts, so tomorrow's practice targets your actual speech. Play it back and note where you stopped and your longest smooth stretch.
Pick one familiar topic, like your job, a recipe, or your daily commute, and keep it for the whole week so every pass repeats the same talk.
Set a timer for 4 minutes and explain the topic aloud while recording. Do not read; talk from memory and accept messy delivery. For example, So the key thing is, start with cold water, because the leaves turn bitter if you pour it straight on is a fine third-pass sentence.
Immediately reset to 3 minutes and deliver the same talk again, then finish with a 2-minute pass. Shrinking time pushes you to retrieve the same ideas faster each round.
This is the self-check: replay the third recording once and mark every place you stopped inside a sentence. It costs about two minutes, and those marks tell you exactly where your fluency breaks down.
Budget roughly 12 minutes total, including playback, and once per day is enough. The repetition across three passes, not daily frequency, carries the exercise.
The rationale for repeating the same talk is planning economy. The first pass spends effort deciding what to say, so later passes spend it on how to say it, which is where smoother phrasing and clearer order tend to appear. Expect less effort, not polish, by pass three. one small classroom study paired timed repetition with rubric self-assessment and reported fluency gains, while other studies found repetition alone did not always improve fluency. Treat the loop as practice with a built-in self-check, not a guaranteed method, and stop if the third pass feels stale; swapping topics weekly is a valid alternative to repeating daily. Mid-clause pauses weigh most heavily on how fluent you sound, and second-language speakers tend to pause inside clauses, so those stops are the right thing to track rather than speaking speed, which research links weakly to fluency.
Record a 90-second voice note today explaining something you know well, like your commute. Set a timer for 4 minutes and explain the topic aloud while recording. Shrinking time pushes you to retrieve the same ideas faster each round.
Worked demonstration: explaining how to make tea
Pick the same topic you used in your loop, making tea, and run one full pass so you can see what a normal, imperfect attempt looks like. This is the 4-minute version, spoken aloud with a timer running, from a learner's real-feeling start.
Here is how the opening might sound, stumbles and all:
Okay, so, making tea. Um, first you need, what's it called, a kettle. And cold water, you should start with cold water because... the leaves? No wait, that comes later. The water heats, and when it boils you pour it, um, over the tea leaves. If the water is too hot it can taste bitter, so, some teas need cooler water.
What happened there: two restarts, one mid-sentence stop, and a moment of going back to fix the order. That is a normal attempt.
The fix-it-out-loud move, no wait, that comes later, is actually useful, because it shows the speaker reordering ideas in real time rather than freezing.
Then the explanation closes with the structure it earned through the mess: So the main steps are, cold water, boil, pour, wait. The waiting time depends on the tea. By the end of the pass, the speaker has a skeleton to reuse in the 3-minute round, which is the point of repeating the same talk. Expect the same pattern in your own first pass: mess first, skeleton second. For the check on this pass, replay the recording and count your mid-sentence stops, using the self-check in the loop above. In this sample that count is one, from the stop at because... the leaves. If your third pass shows fewer stops than your first, that is progress you can hear.