BUILD RECORD

vLLM Video Intelligence

A contact-sheet pipeline made it cheap enough to extract useful data from an entire video library.

Reading one video with a vision model cost about eight cents, which kills anything meant to read a lot of them.

video understanding vision models cost engineering retired

Cost was the call count, so it made fewer calls.

vLLM Video Intelligence

Retired

WHAT IT IS

A video-analysis engine that tiled eight frames into one contact sheet, so cost stopped tracking length.

IMPACT

Reading a video went from about eight cents to under a penny, about a tenth of the cost. That made a free audit of a whole account possible.

SCALE

100 Shorts from one channel, read in seven minutes for about 70¢.

STACK

contact sheets GPT-4.1-mini Gemma 27B Whisper

MY ROLE

Architecture, tiling engine, model comparison, explorer.

VERIFIED

Open source, released with the run's own explorer, every prediction still browsable.

THE EXPLORER

What came out of a video

The explorer: 100 videos, 496 topics, 3.2 hours watched.

THE METHOD

Video became data

video · the path to data
sampled · tiled · read · returned

01 Sampled and discarded

A frame every couple of seconds, near-duplicates dropped.

02 Tiled the survivors

Eight went to an image, turning forty calls into three to five.

03 Read screen and sound

GPT-4.1-mini and Gemma 27B read the sheets; Whisper read the audio.

04 Got data back per video

It returned topics, transcript, and whether the two models agreed.

LIMITS

What it refused to do

• It sampled, so anything between two samples never reached a sheet.

• GPT and Gemma agreed about 85% of the time. Stability, not accuracy.

• One channel of short interviews, not a general rate.

RECORD AND LINEAGE

Where the engine went

STATUS

Retired. Built and open-sourced solo.

SUCCESSOR

FrameIQ, the same sheets across a whole account.

SAME TRICK

Doorbell Intelligence, contact sheets pointed at a doorbell's own recordings.

Back to Side Projects → Read the FrameIQ build →

THE ARCHITECTURE

Architecture map of Video Intelligence Engine: the real components and how they connect
THE ARCHITECTURE MAP

EXAMPLES

The Explorer, Live
THE EXPLORER, LIVE

A RELATED PROBLEM

Is the cost of each model call blocking the product?

This build is retired, but the unit-economics problem remains. If the job is to understand a large video corpus without paying per frame, send it over.

Email me →