Google ads surge, Runway routes media: July 24 brief

Google's ad machine is growing faster as its AI gets better at matching buyers with promotions. That affects any owner paying for search or YouTube exposure. Runway is also handing creative teams a way to control the cost of AI media, while Flux 3 puts video and sound generation inside one model. The common thread is tighter automation.

Google's AI ad engine gains speed

Google advertising revenue reached $81.6 billion in the second quarter, up 14% from a year earlier, according to Alphabet's July 22 results. Search and related revenue rose 17% to $63.3 billion. YouTube ad revenue increased 13% to $11.1 billion.

Those figures matter because they show advertisers have kept spending while Google puts more AI into search and ad delivery. Gemini, Google's AI system, is helping the company understand what a person wants and choose a fitting ad. That can produce better matches. It also moves more decisions away from the person buying the ads.

For a small business, stronger automation makes clean sales tracking more valuable. Google can optimize toward the result you feed it, even when that result is only a weak inquiry or a visit from someone already searching for your name. Our plain-English attribution setup explains how to connect marketing spend to actual customers. Owners using Google's automated campaign type should also review the controls available inside Performance Max.

Runway lets software choose the media model

Runway launched its Media Router on July 23, a system that picks an AI model for each image, video, or audio request. In its official announcement, Runway says a business can set a price ceiling and favor cost, output quality, or speed. The router then chooses from allowed models and reports which one it used.

An AI model is the underlying engine that creates the media. Until now, a team building AI creation into a product often had to select that engine by hand and revisit the choice as prices and quality changed. Runway is turning that research job into a rule that software follows.

This is mainly relevant to companies producing a large volume of media or adding creation features to their own software. A local operator making four videos a month can ignore the plumbing. A team generating thousands of product images should care because a small cost difference repeats on every item.

The caution is control. Automatic selection can reduce effort, but it can also make output less consistent when different engines interpret the same brand instructions differently. Set an approved list, test the same five jobs across it, and keep a human check on customer-facing work. Our small-brand AI video guide covers the simpler end of this decision.

Flux 3 creates video and sound together

Black Forest Labs put Flux 3 Video into early access on July 23. The company says its new system can create clips with sound up to 20 seconds long from written instructions, images, or existing video, according to the Flux 3 launch post. It can also produce dialogue in multiple languages and carry a reference character across shots.

The useful change is that picture, motion, and sound come from one engine. Today, many short AI ads still require separate steps for footage, speech, music, and editing. Combining them can shorten the first-draft process, especially for product demonstrations and social concepts.

Early access means unfinished access for a limited group, not a dependable production release. Black Forest Labs says its published comparisons are preliminary, and image access is due later. Treat the current output as a testing signal. Teams that need a recognizable speaking voice should still apply the consent and consistency checks in our AI voiceover review.

Worth watching

• Google says the Gemini app now has 950 million monthly users. Its next ad format may matter as much as another Search campaign setting.
• Runway's router is live for video, image, and audio. The open question is whether its quality setting keeps a brand's look steady across different engines.
• Flux 3 Image is planned for early access in the coming weeks. Text accuracy inside generated graphics will be the claim to test first.