Build1 publisher2 min readPublished
Runway turns its in-house ad loop into a product that publishes to Meta, Google and TikTok
Runway launched Runway Ads, an ad loop publishing to Meta, Google and TikTok that Runway says lifted the company's weekly output from 77 to about 900 ads. The results come only from Runway's own campaigns, and its learning step runs on data from platforms building rival automation.
The Engineer · Build desk
Drafted by a language model from the sources cited here and checked against its claim ledger before publication. How we use AISend a correction
What happened
- Customers connect an ad account and supply brand guidelines, past ads and product imagery, and the system makes variants, fits them to placements and localizes text, screenshots and voiceover.
- An automated brand check runs before creative goes to human review, and human approval is switched on by default.
- Runway's product page credits performance marketing with more than $100 million in additional annual recurring revenue, without explaining how that figure ties to Ads.
- Runway is piloting the product with select enterprise partners, and its early-access page lays out a generate, approve, publish and learn workflow.
- Omneky already markets a platform that generates creative, connects ad accounts, launches campaigns across major platforms and optimizes on performance data.
Compiled by The EngineerSomething wrong?How this is made
Why it matters
- cost A team that keeps the default human approval and matches Runway's internal pace would be signing off on about 11.7 times as many ads each week as before.
- contradiction Runway's cost-per-subscriber cut and its conversion gain imply two different counts, so neither headline figure can be checked against the other from what was published.
- constraint Every round Runway Ads learns from depends on data supplied by Meta, Google and TikTok, companies building automation that overlaps with the product.
The step that makes Runway Ads an operations product is the last one. Runway says the system reads back performance and generates the next round from what earns spend [1]. That signal comes from the ad platforms, which also carry the ads [17]. Google has been adding creative capabilities to AI Max for Search campaigns, and Meta has described generative AI features for advertisers [16].
Runway built the product for itself first. Anastasis Germanidis, co-CEO alongside Cristobal Valenzuela since February [4], said the company built it because it needed it [2]. Being your own first customer is a good way to find bugs and a poor way to run a control group. Since July 2026, Runway says, weekly ad volume went from 77 to roughly 900 [5]. That is about 11.7 times as many ads [1]. Conversions rose about 34% [6]. If both figures cover the same weeks, average conversions per ad shipped fell to roughly a ninth of the earlier rate [3].
The other figures do not fit one subscriber count. Runway says cost per subscriber fell 41% [6] while spending rose [7]. If cost per subscriber is spend divided by subscribers, that combination needs subscribers to grow by more than 69% [2]. Conversions grew 34% [6]. So conversions and subscribers are different counts, or are measured over different windows [2]. Runway has not published absolute spend, subscriber counts, campaign mix, methodology, pricing or customer-level results [8][10].
For the results to transfer, a customer's campaigns would have to be limited by creative volume to begin with. The customer would also need a baseline measured the way Runway measured its own. The enterprise pilots are where that comparison could come from [3].
The approval design is the part I would copy. A system that localizes voiceover and on-screen text [11] needs a person accountable for each language it writes. Runway lets customers grant automatic publishing per campaign or per variant type, inside budget rules they set [13].
Runway's case against Omneky and the platforms' own tools is integration, with generative video, localization, brand checks and publishing in one system [15]. According to RuntimeWire, buyers will have to judge the creative quality and the effect on campaign economics themselves [15]. Ads follows Media Router, which RuntimeWire reported in July routes generation requests across image, video and audio models [18].
What to watch
- Customer-level results from the enterprise pilots, with a stated baseline and methodology.
- Published pricing for Runway Ads, which sets whether the added variant volume pays for itself.
- Whether Google's AI Max creative features or Meta's generative tools cover the same generate-and-learn loop inside the platforms.