alexandrev-tibco
619d5f921d
- Caption-localized the 5 iPhone 6.7" + 5 iPad 12.9" screenshots into the 5 non-English locales, reusing the en-US framed captures (in-frame UI stays EN). Tooling in fastlane/screenshot_localize/ (Pillow: repaints the pastel caption band, re-renders headline/subtitle in SF Pro; auto-fits + wraps per language). - Rewrote release_notes 1.2.0 in all 6 languages: since 1.1.5 never shipped publicly (last public = 1.1.4), the notes consolidate 1.1.5 + 1.2.0 features. - Fastfile: new `publish` lane — uploads metadata + screenshots for an existing TestFlight build (skip_binary_upload, no submit) so offer codes can be set up before submission. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BkanrydYtrme8wipTzWssG
58 lines
3.1 KiB
Markdown
58 lines
3.1 KiB
Markdown
# Task: generate localized App Store screenshots for MealMood
|
||
|
||
Write `generate.py` (run with `uv run --with Pillow python3 generate.py` from the
|
||
`fastlane/screenshots/_localize/` directory) that produces localized marketing
|
||
screenshots by re-rendering ONLY the top caption text of each base image.
|
||
|
||
## Base images (source of truth — do NOT alter the phone/app area)
|
||
`fastlane/screenshots/en-US/` contains 5 screenshots × 2 devices:
|
||
- iPhone 6.7": `{i}_APP_IPHONE_67_{i}.png` (1320 × 2868)
|
||
- iPad 12.9": `{i}_APP_IPAD_PRO_3GEN_129_{i}.png` (2048 × 2732)
|
||
for i in 0..4.
|
||
|
||
Each image = flat pastel background at top holding a **headline** (bold, black,
|
||
centered, 1–3 lines) + a **subtitle** (regular, black, centered, ~2 lines),
|
||
with a device frame containing an app screenshot BELOW the text.
|
||
|
||
## What to do, per language × per base image
|
||
Languages (output folders, siblings of `en-US/`): `es-ES de-DE fr-FR it pt-BR`.
|
||
Also re-render `en-US` in place is NOT needed — leave en-US untouched.
|
||
Text comes from `translations.json` (index 0..4 matches file index `i`).
|
||
|
||
For each base image:
|
||
1. Read background color = pixel at (10,10).
|
||
2. Detect `phone_top` = first row y in [0.10h, 0.6h] that contains a horizontal
|
||
contiguous run of near-black pixels (r,g,b all < 70) wider than 0.35·w.
|
||
This is the top of the device — the caption band is `[0, phone_top)`.
|
||
3. Repaint the whole band `[0, phone_top)` with the background color
|
||
(this erases the English caption).
|
||
4. Draw the localized headline + subtitle, black, centered horizontally,
|
||
the text block vertically centered within `[0, phone_top)` with a slight
|
||
upward bias (leave ~12% of phone_top as bottom gap before the phone).
|
||
5. Save to `../<lang>/<same_filename>`.
|
||
|
||
## Typography (match the original as closely as possible)
|
||
- Font: SF Pro via `/System/Library/Fonts/SFNS.ttf`
|
||
(variable font — use `font.set_variation_by_name`). Headline = **Bold**,
|
||
subtitle = **Regular**. If a weight name isn't found, print the available
|
||
names once and pick the closest (Bold/Semibold for headline, Regular for sub).
|
||
- iPhone: headline ~72px, subtitle ~50px as a STARTING point.
|
||
iPad: headline ~96px, subtitle ~66px starting point.
|
||
- Auto-fit: if the headline at the start size would exceed 88% of image width
|
||
on its longest line, reduce the headline font size until it fits (min 44px
|
||
iPhone / 60px iPad). Wrap headline to at most 3 lines, subtitle to at most 3
|
||
lines, breaking on spaces. Some localized headlines are long (e.g. the it/fr
|
||
"…in famiglia è facile") — wrapping + shrinking must keep them inside the band
|
||
and never overlap the phone.
|
||
- Line spacing ~1.12×. Gap between headline block and subtitle ~0.5× subtitle
|
||
font size. Colour pure black (0,0,0). Antialiased.
|
||
|
||
## Output & self-check
|
||
- Every output PNG must keep the EXACT same dimensions as its base.
|
||
- After generating, print a table: lang, index, device, headline font px used,
|
||
#headline lines, #subtitle lines, and whether text bottom < phone_top (must be
|
||
true for all — flag any that fail).
|
||
- Do not touch any file outside the five language folders.
|
||
|
||
Total output = 5 langs × 5 indices × 2 devices = 60 images.
|