Reveal preview
The same renderers the read page uses. Nothing here is encrypted or stored — it exists to tune the settings against text you have not memorised.
— judging legibility on text you already know is recall, not reading.
Characters visible is an absolute count, not a share of the line, so every line shows the same amount whatever its length. A line shorter than this cannot be hidden by the window at all — only dithering helps there. Seconds per line is the time the window takes to cross the longest line, once. It is the only control over sweep speed — lines spanned changes how much context is on screen, never how fast it moves. Below about 0.2 the sweep is quick enough for successive frames to read as one image rather than as flashing.
Dither draws only 1/N of the window’s ink each frame and cycles through complementary subsets, so a single captured frame holds a fraction of an already-small window. It relies on the eye fusing successive frames into one form, which holds to roughly 130ms — so 8 frames is the practical ceiling at 60Hz. Raising it lets you widen “visible share” without lighting any more of the message at once. Judge it on text you have not seen: hit Fill with a random credential first, and try actually writing the result down.
Held still, at the ink densities either side of the current dither, using the settings above. The reveal itself can never answer “do these two look alike” — it shows fragments over time and never has both on screen at once.