tds3p-rsvp-reader/TODO.md

TODO

Add anti-aliased reader fonts

The current reader uses TFT_eSPI's bitmap FreeSans, FreeSerif, and FreeMono fonts. They are compact and fast, but their edges remain jagged at large sizes.

Use TFT_eSPI smooth-font (.vlw) assets for the reading view and font-setting previews.

1. Generate or obtain licensed .vlw files for sans-serif, serif, and monospace faces at the seven supported reader sizes: 1, 1.5, 2, 2.5, 3, 3.5, and 4.
2. Place the files in a dedicated SD-card directory, for example /fonts/<style>-<size>.vlw; document these required files alongside the existing /books/<book>/<chapter>.txt layout.
3. Change Board to load the selected smooth font from SD when a font setting is applied, and fall back to the current bitmap font if the file is absent or cannot be loaded.
4. Continue to clear only the union of the old and new word bounds. Smooth glyphs should be drawn with an explicit black background so anti-aliased edge pixels blend correctly without redrawing the status bar or dead space.
5. Measure load time, RAM use, and word-to-word rendering latency on the T-Display-S3-Pro. Keep the cached active font in memory and avoid SD reads during normal word advances.
6. Verify that chapter-title and The End cards use the same selected smooth font and remain aligned with normal words.

Preprocess unfamiliar-word timing

Avoid dictionary lookups while reading. The existing chapter sidecars already carry sparse italic/bold metadata and reserve a timing-modifier field. Extend the desktop preprocessor to populate that field before books are copied to the SD card.

1. Store only marked words: delta-encoded word indexes plus a small extra-delay class (for example, 25%, 50%, or 100%). This keeps metadata sparse rather than recording timing for every token.
2. Have the preprocessor inspect the whole book to flag first appearances of uncommon words, likely proper nouns, non-English terms, and user-supplied vocabulary. Keep language models and dictionaries on the desktop, not the device.
3. At runtime, compare the current word index against the next marked index; do not perform SD or dictionary lookups during word advancement.
4. Combine an extra-delay class with existing punctuation and long-word timing using the strongest applicable delay, rather than multiplying every rule.