Before the check
Chapter Cedar
26:20planned
2 plays of 13:10
Week 1 · day 6 · 30-minute slot
Saturation Listening / Danish
Repeat a phrase
from a Danish story.
I built a listening interface around the moments worth hearing again. Try its actual waveform and timed text with one short, original story.
Play the storyAn original story / A working player
Find the phrase.
Hear it again.
Play the story, follow the words, then choose a passage to repeat. The sound and waveform are the same original sample used in the recordings below.
AI-generated Danish narration. Standard Cedar voice · 15.9 seconds · an original fictional story.
Play the original sample. Enable JavaScript for word highlighting, waveform seeking and phrase loops.
Loading the word and waveform player. You can already play the original sample below.
Read the complete original Danish story
Ved den lille havn fandt en cyklist en blå kop på en bænk. Koppen var tom, men under den lå et kort. På kortet stod der: Kig mod vandet. Da cyklisten løftede blikket, sejlede en lille papirbåd langsomt forbi.
The word timings are reviewed ASR estimates for this sample. They demonstrate the listening interface; they do not measure the original study-text alignment pipeline.
The source interface, recorded
The same small
piece of listening.
These genuine recordings show the original listening interface with this fictional sample. The player above retains its waveform and timed text, with a small static adapter for this page.
What happens in the desktop recording
The 24.23-second recording starts the AI-generated Danish story. A click seeks to a later moment. A mouse drag marks a region around “Koppen var tom, men under den lå et kort.” The interface repeats it, pauses at the region end when looping is disabled, then clears the region and zooms. Keyboard controls are exercised at the end. English action captions describe the workflow; the full original story is available above. Seeking and looping change which parts are heard.
What happens in the touch recording
The 17.88-second silent recording uses actual browser touch input. An ordinary tap moves the playhead without selecting a region. Moving before the hold threshold leaves the selection empty. A 550 ms hold arms selection; dragging then marks the phrase. Playback crosses the selected end and returns twice. Clear removes the region, zoom changes the waveform scale, and the settings drawer switches reading views. The captions describe these visible actions.
A separate local workflow check
Choose what comes next.
The saved Mark complete choice moved the plan from Cedar to Harbor. The app calculated the next chapter’s listening time.
After “Mark complete”
Chapter Harbor
29:45planned
5 plays of 5:57
Week 1 · day 0 · 30-minute slot
Saved choice: Mark complete. Cedar completed; Harbor active. The new plan is approximately 30 minutes.
Actual app in an isolated test copy with an unreleased interface fix. Fictional study history and question; example chapter durations. No audio was played in this recorded plan check.
Six completed days were prefilled. One test completion reached day 7, and the app saved one correct answer to one authored question. This checks the workflow, not learning or retention, and does not use the usual 20-question pool.
Watch the 8-second app check
Recorded in the product
A saved chapter choice changes the next plan
What happens in this recording
Actual isolated app with an unreleased callback-scope correction. Fictional history and question; duration values are fixture metadata. No audio was played.
The complete 8.36-second silent capture shows Cedar’s seeded day-6 plan, the one-question result, an actual Mark complete click and Harbor’s next plan. Original speed and every frame are retained; only the video encoding changed.
Read the plans above on a small screen, or open the recording full size. Read the action captions.
Source 5c925b1, with the unchanged callback moved above an early return in a private copy. The original version failed at this button; this correction is unreleased. A fixture manifest supplies source-regression durations without audio files. The app’s phase, slot and persistence code produced the result shown.
What this demonstration includes
The actual components
The waveform selection, touch threshold, loop boundaries, zoom and HTML-audio-driven timed views come from Saturation Listening’s actual components. Their original TypeScript source is retained and deterministically compiled for this route.
The page adapter supplies one static fixture, in-memory reading settings, scoped keyboard controls and the portfolio’s SVG icons. The keyboard sentence button sends a declared synthetic gesture through the actual waveform handlers; it is not described as trusted pointer input.
Read the public receiptOne original fixture
Original fictional text, AI-generated Cedar speech and real PCM waveform peaks. Independent CPU transcription recovered all 39 source words in order after punctuation and case normalization. Timing estimates were reviewed for this sample, with three declared overrides.
No original study corpus, progress database or server is used. This demonstrates playback and interaction, without a claim about learning outcomes, original PDF/SRT alignment accuracy, physical touch testing or human listening review.
Fixture and source notice · React · React DOM · Scheduler
Source components: 5c925b1de588