smolvlm BROWSER
VISION LAB
Small model. Wide perspective. Local-only edition
An experiment in seeing

A little model.
A little perspective.

Six everyday scenes. Three ways to look at each.
Explore the photographs and their crops, or bring your own world into frame.

Input frameBUNDLED PHOTO · 01 / 06
A ginger cat on a cream sofa beside a houseplant and side table.01 — FULL FRAME
A quiet companionDomestic life
A different crop, a different story.
Camera stays off until you ask.
Frames stay in this tab.
Reference notebook 18PHOTO STUDIES
Written reference descriptions · not model-generated captions

A ginger cat rests on a cream sofa beside a sunlit plant and a small wooden table.

A ginger cat sits upright on the soft cushions of a cream sofa.

Daylight falls on a leafy houseplant and a side table with a book.

Avocado toast and poached eggs share a wooden table with a glass of orange juice.

Poached eggs sit on avocado toast on a ceramic plate.

A glass of orange juice stands beside a folded linen napkin.

A cyclist passes stone buildings and a flower shop on a quiet street.

A cyclist travels along a narrow street lined with old buildings.

Potted plants gather outside a neighborhood flower shop.

An alpine lake reflects mountains between evergreen forest and a rocky shore.

Evergreen trees line the edge of a still mountain lake.

Rocks and wildflowers frame the lake with mountains in the distance.

A laptop and notebook sit on a desk beside a coffee cup, lamp, and small plant.

An open laptop and a notebook occupy one side of a workspace.

A coffee cup, desk lamp, and leafy plant sit together on a desk.

Three people tend raised beds and tomato plants in a community garden.

A gardener kneels beside a raised vegetable bed.

Two gardeners work among tall tomato plants.

The starting collection

6 bundled photographs 3 crops = 18 perspectives
Model unavailable — local runtime and weights are not bundled.
About this local-only edition

The plan calls for SmolVLM-256M inference. This build cannot download the required runtime or model weights, and runtime third-party requests are prohibited. The 18 references are written examples, not AI output. Camera preview works when permission is granted; automatic captioning requires the missing model assets.