Offline color analysis costs hundreds. Upload one daylight selfie for a Spring / Summer / Autumn / Winter verdict, plus makeup, outfit, and hair suggestions. GPT Image 2 is available in the model picker. Free to try.
Welcome to ChatIMG Image Assistant
Upload an image or describe the image you want to generate
Click or drag and drop an image
Supports multiple images — AI will blend them together
Next
Done diagnosing? Keep editing this photo in chat
Try a commute makeup look, a new hair color, or an outfit in your best colors. Free to try.
AI reads skin undertone, hair, iris, and lip color to assign Warm Spring / Cool Summer / Warm Autumn / Cool Winter — with confidence score.
Not just season tags: 8 hex codes that flatter you + 5 that wash you out, ready to copy into shopping filters.
Makes a 9-grid share image with BEFORE/AFTER, STYLE KEYWORDS, and BEST COLORS.
Uploads are used for that single diagnosis, stored in private Cloudflare R2, auto-purged in 30 days, never used for training.
Upload a selfie and get a color card. Another lighting, makeup, or outfit is another image, and each image spends credits.
6 share-ready templates for Xiaohongshu / TikTok / Instagram / Threads — 9-grid, 3-grid, and 9:16 vertical cover.
Time per diagnosis
30s
Average 30 seconds from upload to first card — even on 4G.
Korean studio comparable
₩300,000+
Seoul studios start at ₩300k. You can try this for free.
Languages
4
Native ZH / EN / KO / JA output, all color vocabulary localised.
After you upload a selfie above, the model maps you to one season and returns the full palette.
There's a recurring "is personal color analysis a scam?" post that goes viral in K-beauty / Xiaohongshu communities. The skepticism is fair: $300 for a verdict that 25% disagrees between studios feels off. But "scam" is too strong — color theory has solid foundations (Munsell color science, skin-undertone optics, contrast matching), codified by Suzanne Caygill in the 1980s. The actual scam is (a) studios charging $300+ with marketing fluff, and (b) the "your subtype is fixed for life" overclaim. AI makes the cost zero, subjectivity quantifiable (confidence scores), and lets you re-test endlessly — so it democratizes color theory rather than scamming you. Best path: start free with AI; if 3 selfies under different lighting agree, trust the verdict. Only pay for an in-person session if AI is borderline.
Yes, it can be wrong. chatimg.ai's intra-cluster agreement with Seoul studios is 82%; the other 18% is mostly "neutral skin" boundary types (Warm Summer ⇄ Cool Spring), where the AI ships explicit "neutral-leaning warm / cool" labels plus a 0-100 confidence score and sister-season runner-up. How to validate: (a) shoot 3 selfies under matching lighting and triangulate — 3 matching verdicts at confidence ≥80% is reliable; (b) below 70% confidence, the AI prompts you to re-shoot, fix lighting, or remove makeup; (c) borderline cases can cross-check with the hex-text fallback prompt (self-report skin/eye/hair HEX values for a text verdict). The AI is more honest than human stylists because it admits 71% rather than pretending to be 100% sure.
Dyed hair: yes you can test, but specify your natural hair color HEX in the prompt so AI discounts dye interference. Otherwise the model treats the dye color as your natural tone — wrong verdict. Our playground has a 'natural hair color' field for this. Filters / beauty mode / foundation: avoid these — they contaminate the skin-tone signal and the verdict becomes meaningless. Best practice: natural daylight, bare face (or moisturizer only), undyed hair (or specify dye), 3 selfies from different angles.
Same principle — both analyse undertone (yellow/blue), value, contrast, and chroma. Offline studios drape physical fabric swatches one by one; AI renders all 12 swatches side by side at once. AI wins on cost, speed, and lighting variation. Studios win on tactile fabric and live consultation. We recommend AI first; if you’re unsure, validate in person.
Best: natural light, no filter, bare face or light makeup, undyed hair, face directly to camera, plain background, shoulders-up. Avoid: B&W, beauty filters, coloured light sources (will skew the verdict), sunglasses, large occlusions.
In our internal testing, agreement with Korean studio verdicts is ~82%; the remaining 18% is mostly "neutral" borderline types (Warm Summer ⇄ Cool Spring) where the AI explicitly flags "neutral-leaning warm/cool" with both interpretations. Cross-verify by uploading 3 selfies under different lighting.
Yes. chatimg.ai users own commercial rights to generated images — use them in client reports, social posts, paid course materials. Note GPT-Image-2 output carries an invisible SynthID watermark (compliance, no visual impact).
There is no daily free quota. A new account gets 50 credits once, enough to try one GPT Image 2 image. After that every image spends credits: Plus is 1,500 a month, Pro is 5,000 a month, and top-ups start at $10 and do not expire.