This page collects the questions that repeat most often during real use: whether photos are kept, why outputs change, what makes a better input photo, and why recommendations may differ from expectation.
Keywords
Check whether the photo meets the relevant guide criteria
Confirm that filters and automatic correction are off
Make sure the correction frame matches the actual boundary
Compare repeated runs under matched conditions
Do not confuse recommendation text with the underlying measurement
The tool reads the face as it appears in the uploaded image. Lighting, angle, distance, filters, and obstruction can all move the measured input.
Face shape is sensitive to angle and framing.
Personal color is more sensitive to lighting and automatic correction.
Repeated results under matched conditions are the best trust signal.
Most users run these tools on phones. The bigger issue is usually front-camera enhancement and very close selfies rather than mobile as a platform.
Very close selfies can distort face shape.
Check whether your camera app applies strong automatic enhancement.
Repeat tests on the same device and in the same place when possible.
Style recommendations are an application layer built on top of measurements. It is normal to adjust them according to your styling goal and preference.
The same structure can lead to different suggestions for different goals.
For personal color, a useful tone range can matter more than one exact season word.
The output works better as direction than identity.
No, we do not store or upload your photos to the server. All analysis is performed locally within your browser, and the image data is discarded immediately after the results are calculated.
The structural label can be the same. What should differ more clearly is the styling layer built on top of that structure.
Yes. If you run the analysis after correction, the manual frame becomes the final calculation input.
A warning means interpret with caution, not always that analysis is impossible. But unresolved warnings should lower how strongly you trust the output.