Most analysis was trained on lighter skin and quietly fails on deeper tones. Oyster reads every shade with the same precision, and it is measured to prove it.
The same selfie, read the same way on Fitzpatrick VI as on Fitzpatrick I, so no customer gets a worse answer because of her shade.
Two million real skin images across seven countries, weighted toward deeper tones, so the analysis is just as sharp on the deepest skin as the lightest. That data is the moat, and it grows with every scan.
See the accuracy research ↗
Most skin AI reports one overall accuracy number that hides how it does on Fitzpatrick V and VI. Oyster measures accuracy per skin type and holds the deep end to the same bar as the rest.
How accuracy is measured ↗
Hyperpigmentation, melasma, post-inflammatory marks and uneven tone are the concerns deeper skin brings to the counter most. Oyster measures each one with a number and a place, so the recommendation fits the skin in front of you.
Read: hyperpigmentation on dark skin ↗Accuracy on deeper skin is a claim you can check, not take on faith. Oyster publishes how the engine performs by skin type and welcomes independent review from the clinicians who see these patients every day.
Read Oyster Studies ↗
Held to the same standard on Fitzpatrick VI as Fitzpatrick I, and rising with every scan.
Three gaps that Oyster was built, from the data up, to close.
Trained mostly on lighter skin, so deeper tones sit outside what the model has seen.
Exposure and colour tuned for light skin wash out the very signals that matter on deep skin.
Blind to hyperpigmentation and post-inflammatory marks, the concerns deeper skin brings most.
Bring your own products and customers, run the analysis on the deepest tones you serve, and see the accuracy for yourself before you commit to anything.