Skip to content
Zester
All products

Applied Vision

Style Me Agent

AI hairstyle recommendation that reasons from face shape rather than from whatever is trending.

Domain
Applied Interfaces
Status
Prototype
Stack
Google GeminiVision

7

Face shapes analysed

4

Style directions per result

The problem

Style advice is generic because the input is generic.

Recommendations are made against trends and celebrity reference images, not against the specific face they are meant to suit. The person asking has to translate the advice themselves, which is the hard part.

How it works

  1. 01

    Analyse the actual face

    Google Gemini classifies across seven face shapes from an uploaded reference photo.

  2. 02

    Generate across intent

    Four suggestions are returned — Safe, Trendy, Bold, and Classic — because a single recommendation assumes an appetite the system cannot know.

  3. 03

    Include colour

    Hair colour recommendations accompany the cut, since the two decisions are made together in practice.

Architecture

  1. 01

    Input

    Reference photograph upload

  2. 02

    Analysis

    Google Gemini face shape classification across 7 shapes

  3. 03

    Generation

    Four style directions: Safe, Trendy, Bold, Classic

  4. 04

    Surface

    Style and hair colour recommendations returned together

Capabilities

  • Face shape analysis across seven shapes
  • Four differentiated style directions
  • Hair colour recommendation
  • Reference photo upload

Have an operation this should be pointed at?

Solve a Real Problem