Hi — I'm Yuseong Kim, and I build and run FaceOracle solo from Korea. Product, design, code, writing, and these decisions all come from one person. My background and process live on the editorial team page; this page is the why, in my own words.
Why I only deal with visual elements
Honestly, what bothered me was watching a photo-mood report get mistaken for an appearance score or an identity judgment. The moment you say "AI looks at your face," people start expecting grades, rankings, or your real personality — and I didn't want to build that. A tool that ranks faces as better-or-worse, or claims to read what's inside someone, isn't fun and isn't safe.
So FaceOracle reads only the visual elements that are actually in the photo — color, expression, composition — and turns them into styling ideas and light fun. It runs strictly as entertainment. "This photo gives off this kind of vibe" is as far as I want to go. It's not a scorecard; it just puts the mood of one photo into words.
What this is not
To head off the misreading, let me say it plainly: FaceOracle's results do not judge identity, gender, nationality, health, or personality. It is not face recognition, identity verification, or biometrics.
- Real personality, intelligence, ability, or character — not determined.
- Health conditions, skin issues, or medical age — not diagnosed.
- Sensitive attributes like gender, nationality, race, religion, orientation — not inferred.
- Hiring, admissions, lending, or identity checks — cannot be a basis.
To hold the same line in code and in writing, the features that invite misreading are off by default, and result copy uses vibe language instead of verdicts. The full standards are in the editorial policy and content principles.
So here's how I make it
- Results lead with what's visible — color, expression, composition — in a light tone, not verdicts.
- Celebrities appear only as mood archetypes, never real names. No pointing at anyone, no resemblance claims.
- The original uploaded photo isn't stored — it's discarded right after analysis. Details are in How it works and the Privacy Policy.
- When I change something, I log it with a date in the operator notes below — on the record, not quietly.
Operator’s notes
What the operator changed, by hand, with the date.
- Jun 7, 2026Ads run only on the reading pages
Ads run only on reading pages like the blog, guides, and about — never on the photo-upload screen, the result screen, or screens like deletion and the daily fortune. As an operating rule, the code automatically keeps ad units from ever sitting right next to a photo or a result.
- Jun 7, 2026Took the age-number card out of the default view
The result used to show a guessed age as a number. Since that can read like judging a person, it no longer appears in the default view. The point is to stay on the mood that comes from a photo's color, lighting, expression, and composition.
- Jun 4, 2026Pulled every safety notice into one place
The safety notice used to be worded a little differently across the result screen, guides, and blog. I pulled it into a single source so every surface now shows the exact same sentence — no drift.
- May 30, 2026Swapped ‘look-alike celebrity’ for a mood archetype
I generalized the ‘look-alike celebrity’ card from real names into a mood/style archetype. Pointing at a specific person or asserting a resemblance isn’t what I want this to be. I removed the image-search link at the same time.
- May 23, 2026Reworded the daily fortune rotation copy
Reworded the lines that rotate in the daily fortune so the copy reads as a light one-liner instead of a verdict, and added a small visit-streak badge for people who come back day to day.
- Apr 24, 2026Took the verdicts out of the result wording
Stripped phrasing that sounded like a verdict about age or personality out of the result cards, and rewrote them to describe only what’s visible — color, expression, composition — in a lighter tone. This isn’t a scoring service.
- Apr 12, 2026Turned the misreadable features off by default
Decided to leave features that are easy to misread as judging a person — gender detection, national-look, numeric looks scores — disabled by default. They don’t fit the idea of reading a visual vibe just for fun.
If you spot something to fix
If any wording feels like a verdict or invites the wrong read, tell me — I read it and fix it myself. Questions, corrections, and deletion requests all go to one address: yuseong2099@gmail.com.
Related
- About FaceOracle — what the service makes
- Editorial team — who writes and reviews
- Editorial policy · Content principles