← Back to blog

Skin Tone Accuracy in AI Photography: Why It Matters

8 March 2026 · 6 min read

There's a problem most AI headshot companies don't talk about: their models are biased toward light skin.

Research published in Nature found that 79.6-86.2% of major AI training datasets feature light-skinned individuals. Studies in AI & Society confirm that popular image generators consistently favour lighter features, often lightening skin tones and anglicizing facial features.

Why This Matters

When AI tools produce results that lighten your skin, smooth your features, or make you look "less like yourself," that's not a minor inconvenience — it's a fundamental product failure. Your professional headshot should look like you, not a filtered version of you.

This affects professionals of colour disproportionately. If you're Black, Brown, South Asian, or any non-White ethnicity, most AI headshot tools will subtly (or not so subtly) alter your appearance.

How Per-User Training Solves It

The Crafted Persona uses per-user LoRA fine-tuning. Instead of relying on a general model to approximate your appearance, we train a dedicated AI model on your specific selfies. This model learns your exact skin tone, facial structure, and features before generating any images.

The result: images that look like you in professional settings, not a lighter-skinned version of you.

What This Means in Practice

Representation in professional imagery matters. Your headshot should make you look professional — not make you look like someone else.

See what your headshots would look like

Upload 1 selfie. Get 1 free AI headshot. No payment required.

Try Free Preview

Or take the 7-Second Brand Audit to score your professional image.

See the difference yourself →