
For overseas developers: in mainland China, Android apps are distributed through many phone makers' and platform stores rather than Google Play. Listing on Android in China usually means dealing with Tencent MyApp (应用宝), Xiaomi, OPPO, vivo, Honor, Huawei and other channels at once. Each has slightly different rules for the number, aspect ratio and file size of description images and whether device frames are allowed. If every release means "open Photoshop and Save As six times", an indie team is quickly worn down.
The goal should instead be: one logical master → automated export → channel-level compression and spot checks.
Record these in a spreadsheet or stores.yaml (sample fields; values per each console's current prompts):
| Channel | Description images | Main ratio | Per-image size | Icon size | Orientation rule | Notes |
|---|---|---|---|---|---|---|
| Tencent MyApp | 4–5 | Portrait 1080×1920 common | Often strict | Often strict | Sizes should match | Search carries big weight |
| Xiaomi | ≥3 | Portrait/landscape | Per console | 512 common | Tablet images can be uploaded separately | — |
| OPPO | 3–5 | Portrait 1080×1920 common | Per console | Square corners | — | — |
| vivo | 3–5 | Portrait or landscape | Per console | Size-sensitive | Must be all landscape or all portrait | — |
| Honor | 3–5 | Docs commonly 9:16 / 16:9 | ≤5MB range | Docs commonly ≤200KB | Single orientation | Device frames must match |
| Huawei | 3–8 | Multiple specs | Per console | Special specs | Portrait or landscape (per docs) | Don't mix up with Honor's console |
Write "per console" in the table header: third-party articles go out of date, so the table must be editable.
Define for each frame:
frame_id (e.g. 01_hero)Design rules:
Recommended master artboard: 1080×1920 (9:16). Export matrix:
master/01_hero.png … (no frame, uncompressed)framed/honor/... (only when the channel allows it and the device matches)export/<channel>/... (scaled to the target dimensions and compressed)For landscape channels, create separate 1920×1080 logical frames alongside 1080×1920; don't stretch portrait images.
Suggested naming:
<app>_<version>_<channel>_<frame>_<WxH>.png
For example: moodjournal_1-2-0_vivo_01_hero_1080x1920.png
Compression strategy:
campaign/ directory and switch back to evergreen/ after the campaign, so holiday assets don't linger.Teams of two or fewer can standardize on:
For batch headline application and multi-size export, you can generate the master with AppShot Maker and then feed it into your own compression script. The key: the master never goes to a store; only export does.
Batch production is usually slow because headlines are baked into pixels. A sturdier structure is:
Render the copy layer at export time with a template tool or script. Then "changing just four headlines" doesn't mean recutting six sets of images. Put volatile information like version numbers and campaign dates in the store's "what's new" field whenever possible, rather than burning it into screenshots.
Before replacing a store's images wholesale, keep the previous export/ archive for at least 30 days. If the new images hurt conversion, you can roll back within hours instead of redoing everything overnight. Name archive directories YYYYMMDD_version and record the "reason for change" (new feature / campaign / experiment).
We recommend noting each channel's "last approved icon MD5 / screenshot set ID" in your table, so you don't accidentally upload unapproved drafts when resubmitting.
If your product is also on the App Store, don't mix your China Android batch directories with your iOS ones. iOS has its own size set and review sensibilities; China's Android channels care more about file size and consistent orientation. Only the "logical frame copy" and UI source files are shared. Export device frames, corner radii and status bar styles separately to avoid "one set of images for everywhere" looking bad on both platforms.
Teams that only serve China's Android market can still reserve an empty en_export/ directory, so when they go global they can reuse the same logical-frame story and swap only the language layer.
The cost of multi-channel store images comes from "rule differences", not "design talent". A differences table + logical frames + automatic compression + quick QA turns six redesigns into one design. Treat the console prompts as the final truth and documented numbers only as starting values, and this workflow will keep working in 2026.