On 30 September Runway announced Runway Ads: you connect an ad account and a brand kit, the product generates video and images, publishes the approved variants, reads the results and makes a new round based on what earned spend. For now it is piloting with select enterprise partners.
- Every variant goes through an automated brand check before it reaches human approval.
- Human approval is on by default, and the budget moves within limits the team sets.
- Runway's own ad programme went from 77 to about 900 ads a week since July, per the company.
From 77 to about 900 ads a week. That is Runway's own ad programme since July, per the company.
The number explains the product better than the description. Runway says it outright: in paid advertising volume wins, the more variants you run, the faster you find the ones that work. The bottleneck is rarely analysis. The bottleneck is who makes nine hundred ads.
I have worked with a brand kit that falls apart at the third format. The hero visual looks great, and in the square for the story the logo hits the edge and nobody saw it coming.
Here the machine takes exactly that work: resizing, localisation, brand checks before approval. The dull half of the craft. Good that it takes it.
What bothers me is the other part. The loop closes on what earned spend. The platform spends where an ad works, Runway makes more of the same. A loop that learns only from what earned spend quickly reaches the most effective ad, and that is rarely the best one. That is why human approval, on by default today, is the most important setting in the whole product.