Supplements the main spec and the native-scaling amendment
The annotation body text was specced at 36px (canvas), matching the caption size. This is too small — the manifest is a standalone text section that users read without the visual context of the images above. It needs more presence.
| Element | Old size | New size | Rationale |
|---|---|---|---|
| Annotation body text | 36px | 42px | Primary reading text of the manifest — needs to be comfortable at arm's length |
| Annotation type label | 28px | 28px | No change — metadata stays subordinate |
| Annotation image ref | 28px | 28px | No change |
| Manifest section header | 28px | 32px | Slight bump for visual weight parity with updated body |
| Annotation badge diameter | 64px | 64px | No change |
| Spacing between entries | 48px | 56px | Slightly more air between entries at larger body size |
Caption text (beneath images in the grid) stays at 36px italic — its context is established by proximity to the image, so it can afford to be quieter.
The export background is a warm off-white — not pure white. This is one of the strongest editorial signals in the composition; it makes the export feel like paper and differentiates it from the white backgrounds of messaging apps, docs, and tools where it gets pasted.
The warm cream background combined with white image cards creates a subtle depth — the images sit on the page rather than floating in void. JPG export should use #F4F2ED as the fill, not transparency.
Screenshots bleed to the card edges — no inner padding. The card's border-radius clips the corners of the screenshot. This maximizes the image display area and avoids the "picture frame" effect that padding creates.
Single-image export. The image is always 12-col (full width), regardless of native size or annotation count. The scale ceiling rule is relaxed for single-image exports — the user chose to export one image, and it should fill the composition.
If the single image has 0 annotations, the manifest section is omitted entirely. The export is: title block → image (with optional caption) → footer.
Two-image export. The algorithm assigns spans normally based on priority and scale ceiling:
Three-image export. Natural compositions:
The packing algorithm handles this naturally — no special-casing needed beyond the standard rules.
An export with zero annotations across all images omits the manifest section entirely. The export is: title block → image grid (with optional captions) → footer. The superline still shows "0 ANNOTATIONS" for completeness.
No special handling. The packing algorithm continues filling rows. The export gets taller. For the stitched JPG output, the image simply extends vertically. For PDF and multi-file outputs, see pagination (deferred).
The primary export is a single tall JPG image at 3000px width. This is paste-friendly (Slack, iMessage, email, docs) and requires no special viewer.
PDF export with fonts embedded (Cal Sans, Source Sans 3). Pagination rules are deferred — for v1, render as a single continuous page matching the stitched JPG layout.
Deferred: Multi-page PDF pagination. Open questions: Does the manifest start on its own page? Do page headers repeat the title? Do badge numbers reset per page? Revisit after v1 validates the single-page layout.
Splits the export into separate JPG files. Split points occur between image grid rows — never mid-row. Each file includes the title block as a header and the footer. The manifest appears only in the final file.
Deferred: Exact split-point logic (height threshold? fixed number of rows per file?). Revisit alongside PDF pagination.
| Question | Context |
|---|---|
| Scale floor: 50% or lower? | Very large screenshots (4K captures) may hit 50% at 12-col. Is that too aggressive? |
| 3+6+3 composition | Centered hero with small companions — looks great in theory, needs real-content validation |
| Portrait / extreme portrait cutoff | At what aspect ratio should a portrait image be treated as "extreme" and trigger special handling? |
| PDF / multi-file pagination | Page break rules, header repetition, badge numbering across pages |
| JPG quality vs file size | 92% is a starting point — may need adjustment based on typical screenshot content (text-heavy = needs higher quality) |
| Annotation badge scaling | Should badge size scale with image size, or stay fixed? Fixed is simpler but may look oversized on 3-col images |