TasteCut

BRIEF · V1 · 00:00:00

A brand walks in.
A brief that knows what they love walks out.

TasteCut is an agent for video editors. It reads a brand's audience on Qloo's taste graph — what that audience actually watches, hears and buys, with an affinity score on every name — and cuts it into a working edit brief: pacing, type, grade, music, references. Numbers with evidence, not vibes.

SLATESCENE 01 · TAKE 1
Deliverable
try

How the agent works

  1. 01 READ A small model reads your slate: cleans the brand name, guesses the category, lists proxy brands in case the client is too small for the graph.
  2. 02 RESOLVE GET /search on Qloo finds the brand entity; if it isn't there, the agent grounds through its cultural neighbours instead.
  3. 03 GRAPH Seven parallel /v2/insights calls: demographics, taste tags, artists, films, TV, neighbouring brands, podcasts — each with affinity scores.
  4. 04 DERIVE No model here. Genre tags map to a tempo index → cuts per minute and average shot length; tags pre-score 13 colour grades and 14 type pairings; artist genres set the BPM window.
  5. 05 WRITE Two writers run in parallel on the evidence and return the brief as structured JSON, each section naming the affinities that drove it.
  6. 06 MATCH Your reel is resolved on the graph and compared with the brand across artists, films, TV, brands and taste tags — weighted-Jaccard overlap → a score you can argue with.

Why this only works with Qloo

Ask a language model what Aesop's audience listens to and it will guess from the brand's own marketing. Qloo's taste graph knows what that audience actually over-indexes on — across music, film, TV, podcasts and other brands — with an affinity score for each. TasteCut turns those numbers into craft decisions an editor can act on in the next hour, and the Why Qloo card shows the model-only guess next to the grounded brief so you can see the difference yourself.