Clean company logo wall for print: fetch SVGs via the Wikimedia Commons API and recolour with CSS, instead of scraping a flattened PNG
CONTEXT: a print card needed 11 company logos (OpenAI, Anthropic, Google, Meta, Apple, Palantir, SpaceX, Jane Street, Citadel, DeepMind, Tesla) in a single light-grey tone on a dark background. My first attempt reused the target website's flattened logo-wall PNG with background removal; the founder rejected it as bad quality. Vectors fixed it in one pass.
RECIPE (python stdlib only):
1. Search Commons for titles: https://commons.wikimedia.org/w/api.php?action=query&list=search&srnamespace=6&srsearch=<name> logo svg&format=json with a descriptive User-Agent (anonymous UAs get 403). Filter titles ending in .svg and eyeball the top 5; picks that worked: 'OpenAI logo 2025 (wordmark).svg', 'Anthropic logo.svg', 'Google 2015 logo.svg', 'Meta Platforms Inc. logo.svg', 'Apple logo black.svg', 'Palantir Technologies logo.svg', 'SpaceX-Logo.svg', 'Jane Street Capital Logo.svg', 'Citadel LLC Logo.svg', 'DeepMind new logo.svg', 'Tesla Motors Logo.svg'.
2. Resolve the file URL with prop=imageinfo&iiprop=url and download.
3. Inline each SVG in the HTML: strip the xml/doctype prologue and the width/height attributes (add a viewBox from them if missing), then force one colour with .lg svg *{fill:#E4E4E4 !important; stroke:none !important}. This neutralises Google's multi-colour fills and any inline style attributes.
4. Normalise visual weight by height, not width (svg{height:44px;width:auto;max-width:220px}), and give icon-only marks (Apple) and stacked marks (Jane Street) a larger height so they don't look tiny next to wordmarks.
5. 11 items in a 3-column grid: centre the last two with grid-column spans.
GOTCHA: Apple has no wordmark on Commons; say so rather than faking one. Check each SVG's licence note if the piece is commercial.