Doc-library and ticket-grid tiles: hero labels must survive a second crop—do not shrink an entire legacy .vsd into a square chip
`thumbnail-visio-vsd-jpg` targets attachment lists, portal cards, and board covers. Pipelines resample and sharpen again, so hairlines that were legal in Visio fuse into gray smears. Bold the primary path, enlarge critical numerals, and center the story inside the tile safe zone so rounded masks do not amputate corner callouts; simplify legacy gradients that map to muddy mid-tones.
Thumbnail workflow: measure live tile pixels and corner radii, export JPEG with 2–3× headroom on longest edge, then review in a staging list on non-retina hardware
- Pull design specs or screenshots for tile width, mask radius, and overlay scrims; lay out a thumbnail-specific crop in .vsd instead of uniformly shrinking a full architecture page.
- Drop JPEGs into the pre-prod grid on retina and non-retina displays; if only the lightbox is sharp while the chip is mush, raise source longest edge or simplify layers—do not blame users for not clicking through.
- When thumb and HD assets coexist, separate paths or filenames and log both against the same .vsd source—avoid CDN cache-key collisions where a low-res thumb overwrites the HD URL.
VSD to JPG (thumbnails) FAQ: second crops, rounded masks, dark-theme tiles, retina vs non-retina grids
The same JPEG is crisp in detail view but mush in the list—undersized export or frontend compression?
Investigate both: increase source longest edge for the thumbnail chain and verify the frontend is not over-compressing; acceptance is the grid pixel, not the file opened alone.
Dark-theme lists use charcoal backgrounds—white JPEG tiles glare; CSS filter fix?
Filters often kill hairlines; ship dark-matte JPEG variants or mandate PNG/WebP for themed skins instead of one white JPEG for every chrome.
A milestone badge must show in the corner but vanishes under the mask—Visio padding or component change?
Prefer moving badges inside the published safe zone with bleed in Visio; if the component cannot change short term, export a wider thumb-specific canvas.
Batch thousands of thumbs—outlier mega-JPEGs slow first paint; how to isolate them?
Flag `heavy` rows in manifests, lower quality or split subfigures for those sources, and pair with CDN tiering or lazy loading so one monster file does not drag SLA for the whole grid.
Editors clicked “compress media” and both thumbs and inline figures softened—how to tell upload vs CMS destruction?
Compare origin-object hashes to CMS derivatives; if derivatives are smaller with editor metadata, roll back to the raw JPEG and restrict bulk compress actions.