Index of automated operations The Automation Files

The Product Filmmaker

007
Video Factories · Equipment retail · e-commerce
LIVE
Deployed
July 2026 · first wave published
Format
recurring character presents catalog items
QC
speech scored per render, fail-closed

Product video sells equipment, but filming hundreds of SKUs is economically absurd for a mid-size store. Generation engines change the math.

For each slot, the system picks a product, writes a scene where a recurring character presents it — a talking head with a personality, not a stock avatar — and renders the clip with a state-of-the-art video model. Speech quality is scored automatically; a clip below threshold is regenerated until it passes or the slot is skipped.

Character formats are developed and retired based on performance data the system collects itself. New show, same factory.

EXHIBIT A

Catalog in, clips out

Products flow from the store database into scripted scenes automatically — hundreds of SKUs become a content backlog.

SKU scene script video model speech QC publish
a camera crew made of cron jobs
EXHIBIT B

Speech scored 10/10

Automated speech-quality scoring gates every render; the current character format holds a perfect clarity score.

filmmaker — daily slot
▸ product: chainsaw G-52
▸ scene drafted, 18s dialogue
▸ render #1 — speech 10/10 PASS
▸ thumbnail picked (frame 41)
▸ scheduled 12:30 ✓
perfect diction, every take
EXHIBIT C

Formats as experiments

Characters are A/B experiments with collected metrics — kept, tuned or replaced by the data.

10/10
speech clarity, current format
4h
data-collection cycle per format
the cast answers to analytics
Keep going — next file in the archive The Seven-Brand Grid