How AI matching works
Last updated: Jul 1, 2026
Veluvy’s visual-preference feature is optional, private, and built around consent. It estimates how visually compatible a profile may be with your own personal taste. It is never a rating of how attractive someone is.
It’s completely optional
You can use all of Veluvy, including profiles, search, matching and messaging, without ever turning on visual matching. It only activates when you explicitly opt in, and you can opt out at any time in Settings.
You train on synthetic faces
We show you pairs of AI-generated, synthetic faces, not real people, and ask which you’re more visually drawn to. From your choices, Veluvy builds a private “preference profile” that describes your taste. The same synthetic face pool is reused across users to save bandwidth and cost.
What we show, and what we never show
We show you, and only you, things like:
- “Visual preference compatibility: 78%”
- “This profile appears close to your trained visual preferences”
- “Low confidence. Train more comparisons to improve accuracy”
We will never:
- Give anyone a public or global attractiveness score
- Rank real users by beauty or show how “attractive” someone is
- Infer religion, ethnicity, sexuality or other protected traits from photos
Both people must opt in
A visual-preference compatibility estimate is only ever calculated when both people have opted in and provided a photo for matching. Results are private to the person viewing them and are always framed as personal preference, never objective judgement.
Your data, your control
You can delete your training data, delete photos used for matching, or delete your whole account at any time. See Data Deletion for details.
Questions?
Email [email protected] and we’ll walk you through how it works.