Lens
Flagship demonstration. Human side sees only glyphs. Machine side runs the real decoder. Vision mode discards the text and reconstructs from a rendered image bitmap.
Glyph stream only — no Latin letters in the bitmap.
Waiting…
Direct mode uses the WASM codec when available (JS fallback otherwise). Vision mode rasterizes glyphs with rotate-within-cell, applies blur/noise, then matches cells with multi-angle deskew against a pixel template catalog — plaintext is discarded. Modest rotation often round-trips; large degradation may fail honestly.