Skip to content
Qortex
qortex/for-developers

video → structured.signals

You don’t need another model to train, host or babysit. Qortex runs the video intelligence models; you get back structured, evidence-linked data about brands, content and creative — shaped to drop into the system you already have.

Request API access Read the signal anatomy
Hosted
Models trained, run and maintained by Qortex
Evidence-linked
Signals point at timestamps you can verify
Two layers
Content Intelligence + Creative Intelligence
Signal anatomy

Every signal answers four questions: what, how, where, and what else.

illustrative_signal.json
{
}
Illustrative · not a final schema or endpoint
01 / brand

Brand detection

What was found. A named entity your product can index, filter and join against your own data.

Integration patterns

Three shapes it tends to take.

Conceptual, not prescriptive. Your architecture decides; these are the patterns teams usually start from.

A

Library indexing

Run an existing library through analysis once, store structured results alongside your assets, and query them like any other column.

  1. Archive
  2. Batch analysis
  3. Your index
  4. Search & filter
B

Ingest-time enrichment

Attach signals as video enters your system, so every downstream feature — reporting, targeting, review — starts with context.

  1. New upload
  2. Analysis
  3. Enriched record
  4. Downstream jobs
C

On-demand lookup

Fetch intelligence when a user needs it — a reviewer opens a video, an analyst inspects a placement.

  1. User action
  2. Request
  3. Signals
  4. UI surface
Evaluation protocol

Test it on your video, not ours.

A sample response tells you the shape. Only your corpus tells you the fit.

  1. [1]
    Pick your corpus
    Use video that looks like production — your formats, lengths, languages and edge cases. Not a demo reel.
  2. [2]
    Write down the question
    Which brands, which risks, which creative attributes actually matter to your users? Define ground truth for a sample.
  3. [3]
    Compare against the evidence
    Every signal carries timestamps. Spot-check against the moments, not just the labels.
  4. [4]
    Check fit, not just accuracy
    Does the shape of the output map to your schema, your latency budget and your review workflow?
What this page is

A conceptual picture of the output and how teams integrate it. Field names and samples are illustrative.

What it isn’t

Reference docs, an SDK, or pricing. Those conversations start with your use case — talk to the team and we’ll go through formats, sources and access.

See the sample response on the homepage Working with a PM? Send them this

Bring a sample of your video. Leave with a read on fit.

Tell us what you’re building and what your corpus looks like. We’ll help you scope the signals worth evaluating first.

Request API access