Real grayscale is not a generic desaturate: weights decide who gets brighter
Human vision is more sensitive to green; common BT.601-style weights brighten foliage and slightly darken saturated reds. Naive averaging can collapse skin into concrete walls and flatten logo contrast on infographics. `grayscale-converter` targets predictable ramps for papers, color-blind-friendly previews, and turning UI screenshots into document figures. Printing adds dot gain and ink spread: midtones that look fine on a bright LCD can mud on coated stock, so soft-proof instead of trusting the monitor alone. Heavy JPEG sources amplify block edges and chroma noise that turns into gritty texture once luminance-only. Batch exports should agree on ICC embedding and naming so one half of a campaign is not stripped of profiles while the other keeps them, which skews browser-side gray. Before handoff, sample on phones, desktop browsers, and a physical proof because the same numeric gray can read half a tone apart across OLED versus inkjet. Document the luminance recipe, whether a gray coefficient tag is embedded, and whether downstream color management is allowed so a re-save in another app does not silently swap engines and drift an entire batch.
Grayscale conversion workflow
- On `grayscale-converter`, state the goal—accessibility, print, or sketch—so you know whether to preserve embedded gray metadata.
- Spot-check skin, foliage, red/blue logos, and hairline text for readable separation.
- Export to a new file and keep the color master for edits and legal evidence.