LUNPORTFOLIO 中文
AI Applications

Self-initiated concept

Archive Lens

Organizing images beyond “looks similar,” exploring composition, light, and mood as adjustable retrieval dimensions.

Scope
Research framework, retrieval experience, visual prototype
Duration
Three-week concept study
Tools / Methods
Embedding research · Figma · Visual taxonomy
Year
2026
A large optical lens scanning layered image plates and linking similar fragments with orange nodes
Similarity is not one answer; it depends on the relationship you are looking for.

Retrieval premise

Turn black-box similarity into an adjustable lens

The prototype separates search into composition, color temperature, light direction, material, and mood. People can weight or lock a dimension and watch the archive reorder, making result changes legible.

Three actions in a search

  1. 01

    Point

    Choose a reference

    The system begins with broad similarity without assuming one correct relation.

  2. 02

    Focus

    Weight a dimension

    Increase composition, light, or mood weight and observe the reordering.

  3. 03

    Save

    Save a relation, not a folder copy

    A collection stores the query rule so future images can join automatically.

Continue with Atelier AI to see how comparison and convergence work after generation.

Back to AI Applications