Back to demos
Demo 01 — Photoshoot Curation

91 wedding photos.
AI clusters in seconds.

The pipeline scans every image, generates a CLIP embedding vector, groups visually similar frames into moments, ranks each frame by sharpness and cluster centrality, and surfaces the best selects — all running locally on your machine.

Demo dataset — Smith Wedding 2026
Ceremony
Ceremony
Portraits
Portraits
Reception
Reception
Dance floor
Dance floor
Golden hour
Golden hour
91 total frames
~6 sec demo run
What the pipeline does
01
Scan

Reads every file, parses EXIF metadata (camera, lens, timestamp, GPS).

02
Embed

Passes each image through CLIP ViT-B/32 to produce a 512-dim visual vector.

03
Cluster

Groups vectors by cosine similarity into moments — ceremony, portraits, candids.

04
Score

Ranks frames within each cluster by Laplacian sharpness and distance to centroid.

05
Output

Returns highlighted selects, cluster summaries, and the full session result.

Run on your own shoots
The real pipeline runs locally on NVIDIA hardware — no upload, no cloud.
Request access