Creating YouTube videos with Google AI: how AI tools help

Many marketing teams have the budget and the strategy, yet their YouTube campaigns never gain momentum. The cause is often production: the hero video is months old, the vertical versions for Shorts are stuck in editing, and there are no variants to test.

In this article, we show how Google's AI tools close this gap: Veo and Asset Studio for new footage, ViGenAiR for repurposing existing videos. We also show where human work still matters.

Platforms demand more videos than your production can deliver

Demand for fresh creative grows with every format. YouTube Shorts, in-feed ads and bumper ads each need their own cut, length and pacing. Production capacity, by contrast, barely grows: nobody hires more videographers just because three new ad groups need testing.

This gap between demand and production costs performance. Campaigns running on a single video can't learn which message and which opening sell best. This is where Google's AI tools come in: not as a gimmick for the odd post, but as a new way to set up the production chain.

Your team shifts from making videos to directing them

Until now, video production mostly meant making things: shooting, editing, adapting. With generative tools, AI takes over much of the execution. Your team's role shifts to three tasks:

• Writing prompts: How do I describe the scene I want precisely enough for the AI to deliver it?
• Combining footage: How do I blend AI-generated scenes with real brand assets?
• Selecting based on data: The AI produces many variants; campaign results show which one wins.

How much time this saves depends heavily on the source footage and your approval processes. We don't put much stock in blanket percentages.

Veo and Asset Studio create new footage from text and images

Veo is Google DeepMind's generative video model. It turns a text description or images into moving scenes, handling light, perspective and motion far better than early AI models.

For advertisers, the key point is that Google now offers generative video creation directly in Google Ads, in Asset Studio. Multimodal video creation has been generally available there since August 27, 2026. In a single step, it takes you from storyboard to finished landscape and vertical ads that can run across the entire Demand Gen inventory. Image assets, such as product image variants with different backgrounds, are created there too.

Three uses that pay off for advertisers

Filling in missing scenes: You've filmed an interview with your leadership team, but one statement lacks a matching visual. Stock footage often looks generic, and a dedicated shoot would blow the budget. A short generated scene fills the gap, in the style you want and without licensing fees for third-party footage.

Testing ideas before the shoot: Before you sign off a large production budget, you can generate the spot from text as a rough draft. Mood, pacing and transitions become visible before a single camera rolls. For a leadership decision, that's far more tangible than a storyboard.

Backgrounds for Shorts: With Dream Screen, you can generate backgrounds and short clips in YouTube Shorts from a text prompt. A tour operator, for example, can showcase its destinations against changing backdrops. Availability and supported languages vary by country, so check before you plan.

Good prompts describe the movement in the scene

Many results disappoint because the prompt describes a photo rather than a scene. A video needs direction on movement:

• Too brief: “A red car on the road.”
• Usable: “Wide shot, a red sports car drives fast on a coastal road, the camera tracks it from the side, motion blur in the background, sunset light, high contrast.”

Give director's instructions: camera angle (drone, macro, wide shot), camera movement (pan, zoom, tracking shot) and lighting mood.

ViGenAiR turns one long video into many short variants

While Veo creates new footage, ViGenAiR solves the more common problem: adapting existing material. The name stands for “Video Generation via Ads Recrafting”. It is an open-source tool from the Google Marketing Solutions team that runs in your own Google Cloud project using Gemini on Vertex AI, and it is explicitly not an official Google product.

A familiar situation for many teams: you have a strong landscape video for YouTube, and you need versions for Shorts and other vertical placements. Until now, that meant an editor reviewing the video, picking the best moments, cropping the frame, redoing the subtitles and making sure the logo stays visible in vertical format.

Smart cropping

Simple tools just crop the center of the frame. If the main person is standing at the right edge, all you see in vertical format is their shoulder. ViGenAiR continuously shifts the crop to follow the most important element in the frame, such as the speaker or the product. Alongside vertical, landscape and square, it has also supported 3:4 and 4:3 since December 2025. Where a frame can't be cropped sensibly, it fills the edges with a blurred, enlarged version of the background.

Short versions by topic

Gemini analyzes the video's visuals and audio and splits it into meaningful segments, so cuts don't land mid-sentence or mid-scene. Using prompts, a target length and your choice of segments, you control which short versions get created, for example several Shorts about value for money from a single long product video. ViGenAiR also generates copy and still images as additional ad assets for Demand Gen.

One long video becomes creative for many weeks.

Variants for testing

Which opening works better: an emotional scene, a product close-up or a text overlay with a key fact? From the same source footage, you get variants with different openings. You test them against each other in Google Ads, and the campaign shifts budget to the variant that performs better. What used to take a day in the edit suite now takes minutes.

AI creates, we review and decide

We never rely on AI unchecked for generated or recut videos. It doesn't always grasp irony or the subtle nuances of a brand message, and it sometimes cuts in the wrong place. Before anything is approved, we check:

• Message: Does the statement still hold when the video is shortened?
• Brand: Are the logo, colors and tone of voice intact?
• Rights and labeling: Are the music and the people on screen cleared for the new version, and does AI-generated content need to be labeled?

The review takes a fraction of the time editing used to. But it decides whether the extra variants actually benefit your brand.

Conclusion: more variants only pay off with rigorous selection

Veo, Asset Studio and ViGenAiR can close the gap between what the platforms demand and what your production can deliver. Campaigns get enough creative to learn which message sells.

But volume is not what makes the difference. When all advertisers use the same tools, the idea, real brand material and selection based on results decide. We compare the variants against the campaign goals and expand the approaches that perform better under comparable conditions. More on this on our YouTube Ads page.