ChatGPT as your makeup artist: analyze your look from a photo — hair, makeup, skin

01 — The trendWhat can ChatGPT's "appearance analysis" actually do?

OpenAI shipped a strong image-generation update, and the "appearance analysis" trend caught on for a reason: the accuracy genuinely surprises. The logic is simple — you upload your photo and ask the model to build a visual board: analysis of your features, what suits you, what doesn't, and concrete recommendations.

The output isn't a wall of text but a laid-out mood board: face proportions, best hairstyles, what to avoid, a color palette, capsule looks. Below are four prompts that build it. One rule for all: attach clean photos with no makeup and no filters, or the analysis goes off.

02 — HairPrompt for hair and face-shape analysis

The first and most striking prompt analyzes the face and matches hairstyles with structure and a harmony level. Copy it as is (it's in English — the model is more accurate that way):

Center the natural portrait (no retouching, real skin, same face).
Add analysis: face proportions, bone structure, key features, symmetry + hair (density, texture, movement). Define balance: where to add/soften volume.
Show haircuts as structure: layers, blunt/U/V shapes, face-framing, bangs (or none), bob, shag, solid vs textured, different lengths. Each: clear construction + short note (elongates/widens/softens) + harmony level (high/medium/low).
Then styling: waves, sleek, blowout, buns, ponytails, etc. — show how styling supports or breaks the haircut.
Add: visual mistakes/effects (heaviness, flatness, imbalance) + why, color (warm + alternative tones, effect on structure), styling principles (volume, movement, parting, finish).
Final: "you in different looks" (capsule styles) + clear direction (best haircut type, styling, color).

03 — MakeupMakeup prompts: what suits you and how to apply it

Next, two makeup prompts. The first asks for recommendations (what makeup suits your face), the second for a step-by-step application scheme. Both require a clean photo with no makeup and no filters.

The application-scheme prompt is the most practical — it draws where to put contour, highlight, blush and brows:

Create a full scheme on where to apply for the contour, highlight, blush and brows to get that result, and the same for the eye, if we want to make a medium eye makeup, where and how to apply to get the amazing result that we have here.

First ask for recommendations, then for the application scheme of the option you chose. That way you get not an abstract "nude suits you" but a map: where and what to apply to repeat the result.

04 — SkinPrompt for skin analysis and skincare

The fourth prompt is a visual skin analysis: type, problem areas and what to do about them (skincare, routine, tools). Again — a clean photo with no makeup and no filters.

Create visual analysis of my skin, my skin type, what problems do I have in which areas, and what to do with that to fix it — skincare, routine, biohacker tools and etc.

It's a fast, honest start, not a diagnosis. Serious skin issues — see a dermatologist; AI is good as a first map, not as a doctor.— Anjela Petkova

05 — How to use itHow do you get an accurate result, not mush?

Accuracy depends on the input — as always with AI. A few rules so the board lands instead of going generic:

• A clean photo, no makeup, no filters — otherwise the model analyzes the filter, not you.
• Good even light, a front-facing portrait — for hair you can add a full-length shot.
• Prompts in English — the model is more accurate for this scenario (as in the example).
• Treat it as a recommendation, not a verdict: the board shows direction, the decision is yours.

And remember: ChatGPT here isn't a replacement for a makeup artist or dermatologist, but an accessible first step that used to cost a consultation.

06 — Why it worksWhy does a model pull off this kind of analysis at all?

The reason is the same as with everything else in AI: the quality of the answer is set by the quality of the input. A clear photo plus a structural prompt that asks not "rate me" but for concrete axes (proportions, bone structure, volume balance, harmony of options) — and the model produces a systematic breakdown rather than a compliment.