Sticker APNG: feathered alpha reads differently on dark vs light bubbles, and sticker stores cap per-file and total pack KB
`sticker-gif-apng` upgrades emoji-like packs from GIF dither to smooth alpha, but sticker marketplaces enforce per-sticker and total bundle sizes. High frame counts slow keyboard cold start. Tier chat previews vs store hero art, re-validate halos on black and white mats, and check whether review wants per-frame PNG exports that must be generated from the same source as the APNG.
Sticker variant: lock canvas and safe margins, optimize inter-frame deltas, binary-search compression to the store byte ceiling
- Read platform canvas, safe margins, and alpha rules; trim GIF-era letterboxing that encodes useless transparent pixels.
- Test on light-gray and dark-gray chat skins; cap peak brightness on flashing stickers for OLED complaints.
- When over KB limits, merge duplicate idle frames before dropping fps, then resolution; log parameters so re-approval is reproducible, not artisanal.
Sticker FAQ: size caps, halos, keyboard performance
Store rejects the animation for size—should we shrink resolution or frame rate first?
Count frames first—long static holds should merge; lower fps before chopping resolution so 3× screens stay crisp. Keep a high-quality master in DAM and ship a compliant derivative.
Edges look gray on dark bubbles but fine on light—matte issue or blend order against the chat background?
Use the vendor’s dark/light preview fixtures; if only dark fails, tune export matting or reduce outer feather. Do not sign off on white-background comps alone.
Keyboard launch regressed after an APNG pack—prove decode vs network to data teams how?
Compare cold start with offline bundles vs CDN pulls; profile decode threads in Instruments/Android Studio. Offline slowness points to files; online-only slowness points to TLS/CDN.
Telegram and WeChat specs differ—can one APNG source drive both?
Batch export per spec from one master with fixed anchors—never stretch one binary. Each size needs its own KB and halo pass.
Review wants per-frame PNGs that might drift from the live APNG—how do we avoid “bait and switch”?
Automate frame extraction from the same build artifact as production APNG and version the script in the ticket. Ban hand-exported review packs that diverge from download URLs customers see.