Version of 15/09/2026

How we use AI

Where artificial intelligence works in Svoista, how we mark it, and what we do not do with your data.

#de

1. Where AI is used

  • Virtual try-on: looks are generated by Google models (Vertex AI and Gemini).
  • AI model: the model image is generated from your photos by Google Gemini, which also checks whether the photos are suitable.
  • Checking looks: before it is shown, every look is automatically checked by a Google model for unacceptable content.
  • Search by photo: pieces in the photo are recognised by Google Gemini.
  • Wardrobe: the type, colour and style of a piece in a photo are recognised by Anthropic Claude or Google Gemini.
  • Search: a query written in your own words is interpreted by Google Gemini, with Anthropic Claude as a fallback.
  • Catalogue: attributes of brands' pieces are tagged by Anthropic and Google models.
  • Support: a draft reply may be prepared by Google Gemini; a person sends it.

Recommendations are not generated by AI: the order of pieces is calculated as how closely a piece's style matches your taste profile (see “How ranking works”).

#poznachky

2. How we mark it

Every look carries the label “AI-generated image”, and the look's file on our servers contains metadata about its origin. When you share a look from the app, we add a caption about AI wherever the phone's system allows it.

Next to recognition results, the app explains that they were produced with AI.

#obmezhennya

3. Limitations

A look is not a photograph: the fit, colour and size of the real piece may differ. Recognition can be wrong. If a result is unacceptable or clearly wrong, tell us through support.