Spoken English Practice
Lord, open my lips. My mouth shall declare your praise.
Psalm 51:15
What we do
A great many new immigrants and missionaries are stuck in the same place: they can read it and they can follow it, but their mouth will not open. That is not missing knowledge, it is an unpractised mouth. This platform does one thing: it lets you say the same words until your mouth remembers them for you. An English sentence on screen, a model recording first, then you read it back; when the microphone hears you stop, the count goes up by one. No score, no red cross, and nobody telling you your pronunciation rated 72. There are two paths now: Scripture sentences read forty-nine times each, or the 100LS route, where you pick an episode you actually like and watch it through a hundred times.
Two paths
Both ways of practising exist side by side and you switch whenever you like; nothing forces a choice. A summary line at the top shows where your time actually went, because seven Scripture sentences runs forty-five minutes and up, one 100LS pass runs twenty minutes and up, and wanting both means clearing seventy minutes in a day.
Ninety-eight verses in a fixed queue, seven a day. One model recording, seven reps to a round, seven rounds, forty-nine in all. The material is settled so all you do is read.
Visit PlatformPaste a YouTube link into your library, then watch that episode a hundred times. You choose the show. Subtitles for the first three passes, then off, shadowing the actors from the fourth.
Visit PlatformThe two paths keep separate records and share only an anonymous id. That is what makes it possible to ask whether the verses still get read once video is on the table, and answering it never required mixing both sets of rows into one table.
The method
Seven reps a round times seven rounds is forty-nine, at seven sentences a day. Those numbers are there to be tested rather than asserted, so reps per round, rounds, and sentences per day are all sliders you can move.
Every round opens by replaying the model recording. Over forty-nine reps a voice drifts, so recalibrating every seven keeps you imitating the reference rather than your own last attempt.
The microphone decides when a rep is finished, so you never reach for a button. Speak, pause, count. A hand hovering over the keyboard breaks the rhythm.
Seven sentences runs about forty-five to fifty-five minutes, and not finishing in one sitting is the normal case. Today's progress accumulates across sessions, so an evening can pick up where the afternoon stopped.
One recording per round, not forty-nine files. Nobody opens a list of forty-nine items, and keeping them all runs to hundreds of megabytes for a single sentence.
The material
Fourteen days at seven sentences a day comes to ninety-eight, every one picked by hand from Scripture and then ordered so that short and long sentences alternate. That way you meet the whole range in the first few days instead of coasting through a week of short ones and hitting a wall in the next. The text is the World English Bible, which is public domain, and each model recording is generated ahead of time. The queue is fixed by the system with no chapter picker: the moment people choose their own material, difficulty becomes something they selected.
Below ninety-eight, sentences start repeating in the second week, and the count of how many times something was read is contaminated by familiarity.
100LS video shadowing
The method comes from Chang Tong-wan's account of teaching himself English in a hundred days. A "pass" means one whole episode watched end to end, never a clip. The first three are a different kind of work from the rest:
No subtitles at all. Just watch it through once.
Turn them on and get the story straight in your head.
Switch to English and write down the words and turns of phrase you did not know.
No subtitles, saying the lines out loud with the actors and imitating their pace and intonation. These ninety-seven passes are the method itself.
Subtitles are switched in YouTube's own menu; the page prompts you rather than doing it for you. An embedded player can only suggest a subtitle track, and an unreliable switch is worse than none.
War, science fiction and fantasy all work badly: thin dialogue and too much vocabulary you will never use. A dialogue-heavy sitcom or romance is the best material, and a single episode of around twenty minutes is the easiest thing to keep up with. Twenty minutes a pass is a little over thirty hours for a hundred; a two-hour film is over two hundred hours for the same hundred, and most people stop before the tenth. The thing that matters most is still that you actually like it.
The library stores the eleven-character video id from your link and caches no content at all. If a video goes away, point the entry at a new id and the passes you have already put in stay where they are.
What counts as a pass
The usual way this method fails is that the passes quietly turn into background noise. If all you count is how many times something finished playing, pass sixty and pass sixty-with-you-in-the-kitchen look identical in the records, and the whole set of data answers nothing. So a pass only counts when all of this holds:
Measured as the union of the intervals you actually watched, not as elapsed time added up. Looping one stretch twenty times counts as that stretch once, and scrubbing back never double-counts. Drilling one line is not the same as watching the episode.
The microphone totals how long you were actually speaking, and the ratio has to clear a threshold before the pass counts as shadowing. The first three passes are listening and note-taking, so this is not asked of them.
Five seconds of silence before you begin. On speakers the microphone records the actors as if they were you, so that pass does not count towards the hundred and the page tells you to put headphones on and start again.
A pass that falls short is still saved, simply marked as not counting, and shown as a hollow square in the library grid. "Watched today but did not speak" is a record worth having, and it should not be thrown away just because it does not count.
Deliberately absent
Pronunciation scoring is not accurate enough yet, and an inaccurate score mostly makes people quit early. This version only counts reps.
Open it and practise. No sign-up, no password, no email address collected.
Recordings live in your own browser and are never uploaded. You can clear them whenever you like.
Playback runs on the official YouTube player, never covered, never stripped of its controls, never downloaded. All we store is the eleven-character video id from your link.
Both paths are live and you can start straight away. Headphones are strongly recommended: on speakers the microphone hears the model recording and the actors, and may count them as you.
Microphone permission required · Recordings never leave your browser