Doc stills to JPEG: one numbered step maps to one frame—never grab a “close enough” transition frame
`animation-still-jpg` powers help centers, runbooks, and ticket screenshots: step N must match UI state pixel-for-pixel. Wrong frames miss modals or pressed buttons and readers stall mid-flow. Multi-step flows should ship numbered JPEG sets, not one canvas crammed with states. Print paths need ~300 dpi and hairline legibility.
`animation-still-jpg`: drive extraction from the step table, then sign off in light/dark themes and print PDF
- Align with the author’s step table: each step maps to a GIF frame index or timecode—no “close enough” frames; redact sensitive UI in the GIF before export.
- Review stills in both light and dark help themes; ensure arrows and step numbers stay readable below ~12 px on phones.
- Bind `step03.jpg` hashes to the parent GIF in Git or a DAM so doc edits force image updates—avoid text changes without new pixels.
Documentation still Q&A: missed modals, blurry print, GIF tutorial out of sync
Three doc images are adjacent frames from one GIF and readers cannot tell them apart—how do we force more information per still?
Pick frames by UI state change, not temporal adjacency—default vs pressed, before vs after response. Add a half-second dwell frame in the source GIF for documentation instead of grabbing mid-transition blends.
Fine gray guides vanish in dark help theme—is that a design token bug or JPEG quality too low?
Validate contrast on the real dark background; if strokes are under one physical pixel, thicken or recolor. Low JPEG quality exaggerates blocking that reads like “broken lines”—raise quality or supersample then downscale.
Embedded JPEG in PDF prints soft—is dpi too low or did the PDF pipeline recompress?
Target paper width × 300 dpi when exporting; disable extra JPEG passes inside the PDF generator. GIF screen captures rarely meet print dpi—plan pixels up front.
We reuse one screenshot set across locales but button copy changed—re-render GIF or edit text only?
If copy lives inside pixels you must export per language and refresh the mapping—never change prose alone. Long term, automate UI captures from localized builds.
Users say “your steps skip one”—was it a wrong frame or a paragraph reorder?
Cross-check step numbers, GIF timecode, and JPEG filenames; walk the flow blindfolded-style during spot checks. Issues usually come from doc reordering without new art, not the converter.