Gemini Omni in Google Ads Asset Studio: Generate Brand New Video Ads From a Brief

If you manage Google Ads for a small or mid-size brand, you have probably noticed the platform quietly filling up with buttons that all say some version of "generate video with AI." It is easy to assume they are all the same feature wearing a different label, and that assumption is costing advertisers real time and real confusion right now. Google shipped two genuinely different AI video capabilities into Google Ads within a few months of each other in 2026, and they solve opposite problems.

One, which we covered in detail in our guide to Performance Max's generative video resizing, takes a video ad you already made and automatically crops or reframes it into aspect ratios you are missing. The other, the subject of this post, does not touch an existing video at all. Gemini Omni inside Google Ads Asset Studio builds an entirely new video ad from scratch, starting from a text brief and whatever brand assets you point it at. Google previewed it at Google Marketing Live on May 20, 2026, and per Google's own accelerate-announcements page, dated August 26, 2026, it is now rolling out more broadly as a free feature at no additional cost inside Asset Studio.

This post walks through exactly how the four-step workflow works, what actually happens when your brand assets are thin or inconsistent, a full worked example for a small ecommerce brand, and the specific mistakes advertisers are already making with it. We will also put it side by side with the older PMax resizing feature in a real comparison table, since the two are genuinely easy to confuse by name alone. If you are building out a broader content pipeline around whatever comes out of Asset Studio, Miraflow's Cinematic AI Video Generator and AI Image Generator are worth knowing about too, for reasons this post gets into later.

What Gemini Omni in Asset Studio Actually Is

Asset Studio itself is not new. It has lived inside Google Ads for a while as the place where advertisers upload, generate, and manage creative assets across campaign types, and it was already built on Gemini, Veo, and Nano Banana models before this update, mostly for image generation, copy variations, and basic video editing tasks like trimming and adding voiceover to a clip you already had. What changed at Google Marketing Live is the addition of Omni, described by Google as its newest and most capable multimodal model, layered into Asset Studio specifically to unify image, text, and full video generation inside one workspace instead of routing different asset types through different sub-tools.

The distinction that matters most for a marketer evaluating this is simple to state and easy to forget in practice: Omni does not edit or extend a video you already shot. It generates the video itself, as a new asset, reasoning from a creative brief and your brand's existing materials about what the ad should look like, sound like, and say. Google's own framing on its announcement page is direct about why this matters: "Creative variety is the biggest driver of ad performance across Google and YouTube campaigns, including Demand Gen." That line is doing real work in Google's pitch. Performance Max, Demand Gen, and YouTube campaigns all serve across channels with very different creative needs, and the single biggest reason smaller advertisers underperform on those surfaces has rarely been targeting settings. It has been simply not having enough distinct, well-made video variants to feed the algorithm in the first place.

Gemini Omni is Google's answer to that specific gap, and it targets a specific kind of advertiser: one who has a product, a landing page, and maybe a handful of static photos, but no video production budget and no in-house editor. Before this, that advertiser's realistic options for getting video into a Performance Max or Demand Gen asset group were paying an agency, doing it themselves with consumer editing software and stock footage, or running with no video at all and accepting the coverage penalty. Omni is a genuine fourth option, not just a faster version of an option that already existed. It also arrives alongside Google's broader push to give advertisers a single conversational entry point into account management, the same underlying direction we covered in our look at Google's Ask Advisor agent, which plans and reasons across your account but still does not touch actual creative production the way Omni does.

Search Engine Land's coverage of the rollout makes a point worth repeating here rather than glossing over: advertisers "will still need to scrutinize whether AI-generated videos genuinely meet their brand standards rather than relying solely on Gemini Omni's interpretation of their guidelines." That is not a knock against the tool so much as a fair description of what any generative brand tool requires from the person using it, and it is the same discipline this post keeps coming back to in the mistakes section below.

The Four-Step Workflow, and Where Each Step Actually Breaks

Google describes the Asset Studio workflow as four steps: Anchor Your Brand, Generate Concepts, Refine with Prompts, and Package & Deploy. That reads clean on a keynote slide, but each step has real mechanics and real failure modes worth understanding before you start generating, especially the first one, since everything downstream depends on it.

Step 1: Anchor Your Brand

This is where you import brand guidelines and a URL, typically your landing page or homepage, into Asset Studio. Behind the scenes, the model is not just grabbing your logo file. It is parsing the page for color usage, typography cues, product imagery, tone of voice in your copy, and whatever structured brand information it can pull, then building an internal working model of what this brand looks and sounds like, one that every later step references.

This step is also exactly where things go wrong for a specific kind of advertiser, and it is worth being honest about it rather than pretending brand anchoring works equally well for everyone. If your landing page has weak or inconsistent brand assets, meaning your product photos vary wildly in lighting and background, your color palette shifts between pages, your logo appears in three different treatments across the site, or your copy reads nothing like how you actually talk to customers in ads, Omni has a much harder anchoring problem to solve. It cannot infer a coherent brand identity that does not exist yet on the page you pointed it at. The practical result in that case is usually generic-feeling output: safe color choices, stock-photography-adjacent framing, a voiceover tone that reads as competent but interchangeable with any other brand in your category.

The fix is not complicated but it is real work. Before anchoring, pick the single most visually consistent, most on-brand page you have, often a flagship product page rather than a cluttered homepage, and consider supplementing the URL import with a short written brand guideline document uploaded directly, covering your actual brand colors as hex codes, your preferred tone in two or three sentences, and two or three reference images that represent the look you actually want. Advertisers who skip this and just point Omni at a messy, inconsistent site are the ones most likely to end up treating the first storyboard as final, a mistake covered in detail later in this post, because a weak anchor tends to produce safe, forgettable output that still looks plausible enough to approve without a second thought.

Step 2: Generate Concepts

Once the brand is anchored, you either type a creative brief in plain language, something like "a 15-second ad for our ceramic mug line, warm morning kitchen setting, focus on the glaze texture," or you select existing static assets, product photos you already have, and let Omni build video concepts around them instead of starting from a blank prompt. Google's material describes this step as producing full video storyboards and motion scenes, not a single finished clip. That framing matters: the intended output at this stage is a set of directional concepts to react to, closer to a rough cut or an animatic than a polished ad ready to spend against.

Google's stated goal here is what it calls reasoning-backed generation, meaning the model is meant to make scene, pacing, and framing decisions that stay consistent with the brand anchor rather than generating visually impressive but off-brand footage. In practice, this is the step where the quality of your Step 1 anchor shows up most clearly. A well-anchored brand tends to get concepts that already feel usable with light editing. A poorly anchored one tends to get concepts that look fine in isolation but do not obviously belong to that brand specifically, the kind of ad you could swap another company's logo onto without anyone noticing.

Step 3: Refine with Prompts

This is the step that separates Omni from a one-shot generator you either accept or discard. Once you have a storyboard or concept worth building on, you can use natural language to adjust individual scenes, swap or rewrite the voiceover script, change pacing (tighten a slow open, extend a beat that needs more time to land), and adjust aspect ratio, all through conversational prompts rather than a traditional timeline editor. Google frames this as real-time editing, and it is the mechanism that turns a rough concept into something you would actually approve for spend.

The honest limitation worth naming is that natural language editing of video is still less precise than a human editor nudging a clip frame by frame in a timeline. If a specific shot needs the product held at an exact angle to show a feature clearly, or a voiceover needs to land a specific word right as a visual beat hits, you may need several rounds of prompted refinement to get there, and there is a real chance you land close but not exact. Budget for iteration here rather than expecting the first refinement pass to nail it, the same way you would not expect a first draft of ad copy to be your final headline.

Step 4: Package & Deploy

Once you are satisfied with a concept, Asset Studio exports it as multi-format assets, generating both horizontal 16:9 and vertical 9:16 variants optimized for different Google and YouTube surfaces, and lets you deploy directly into Demand Gen, Performance Max, or YouTube campaigns without a separate upload step. This is a genuinely useful piece of production friction removed. Historically, getting one finished video into three different campaign types in the right shapes meant exporting, re-encoding, and manually uploading into each campaign's asset group separately. Package & Deploy collapses that into a single action from inside the same tool that generated the creative in the first place.

A Worked Example: A Small Ecommerce Brand's 15-Second Vertical Ad

Here is what the workflow actually looks like end to end for an advertiser who has never touched Asset Studio before. Picture a small ecommerce brand selling handmade ceramic mugs, one flagship product page, a decent but not professional product photo of their bestselling glaze color, and no video assets at all.

Starting materials: one product photo (a mug on a wooden counter, natural window light, shot on a phone but well lit), and a one-line creative brief: "warm, cozy morning coffee moment, focus on the glaze texture and handmade feel, 15 seconds, vertical, for Demand Gen."

Step 1, anchoring: the brand owner imports their product page URL, which happens to be their most consistent page, better-lit photos, a consistent warm color grade, clear copy about the glaze being hand-dipped. They also upload a short note: brand colors are terracotta and cream, tone is warm and unpretentious, avoid anything that looks like a big-box retail ad.

Step 2, generating concepts: they select the existing product photo rather than typing a brief from nothing, since the workflow supports using existing static assets as the seed for a video concept. Omni returns a storyboard: an opening shot of steam rising from the mug, a slow pan across the glaze texture, a hand wrapping around the mug, and a closing shot with the product centered against a softly blurred kitchen background, with a suggested voiceover script reading something like "Made by hand. Held every morning."

Step 3, refining: the first pass is close but generic. The pacing on the opening steam shot runs too long for a 15-second Demand Gen ad, where the first couple of seconds need to earn attention fast. The brand owner prompts: "tighten the opening steam shot to under two seconds, start with the hand reaching for the mug instead." A second prompt adjusts the voiceover: "make the tone a little more conversational, less like an ad script." After two or three rounds of this kind of prompted adjustment, the storyboard reads as something that could plausibly be this specific brand rather than a generic coffee-product ad that any mug seller could have generated.

Step 4, packaging: they export both the 9:16 vertical cut for Demand Gen and, since it costs nothing extra, a 16:9 horizontal cut in case they want to test it on standard YouTube in-stream inventory too, deploying the vertical version directly into a live Demand Gen campaign.

The part worth sitting with here is what this replaces. Before Omni, this exact brand's realistic path to getting a video asset into that Demand Gen campaign was either skipping video entirely, which limits how well Demand Gen can serve into video-heavy placements, or hiring someone to shoot and edit a video from one phone photo and a vague creative direction, a cost most single-SKU ceramics shops are never going to justify for one campaign. The gap Omni closes is specifically for advertisers in exactly this position: real product, real brand, no production budget, and previously no realistic path to a usable video asset at all.

Gemini Omni vs PMax Resizing vs Uploading Your Own Video

Because these two Google Ads AI video features launched close together and both live under the broader Asset Studio and Performance Max umbrella, advertisers keep asking which one they should be using, when the honest answer is usually both, for different jobs. Here is a direct comparison across the dimensions that actually matter when you are deciding which tool fits a given creative gap.

Read that table as a decision tree rather than a ranking. If you have zero video assets and a product photo, Omni is your starting point, since PMax's resizing feature has nothing to resize without an existing video. If you already have one solid horizontal video and are just missing a vertical or square cut of that exact same ad, PMax's automatic resizing (on by default as of the August 17, 2026 rollout we covered separately) is the faster, cheaper path, since it works from what you already have rather than generating something new. And if your category has strict creative approval requirements, a regulated claim, a franchise brand guideline, a legal disclaimer that has to appear in an exact spot, manually uploading your own video, produced or reviewed by a human before it ever reaches a campaign, remains the right call regardless of how good either AI tool gets.

SynthID Watermarking and What It Means for Reuse

Every video Gemini Omni generates carries SynthID watermarking baked directly into the asset, an imperceptible signal embedded in the pixels themselves rather than a visible logo or bug in the corner. Google's stated purpose is provenance and transparency: SynthID lets a video be identified as AI-generated even after compression, cropping, or minor edits, which matters for platform trust and disclosure requirements as synthetic video becomes more common in advertising generally.

This has a practical implication advertisers should think through before they get excited about repurposing an Asset Studio video beyond the campaign it was made for. If you take a Gemini Omni ad and reuse the same clip as organic content on your brand's own YouTube channel, Instagram, or TikTok, the SynthID marker travels with it. That is not necessarily a problem, disclosure of AI-generated content is increasingly standard and, in some cases, a platform requirement rather than an optional courtesy, but it does mean you should treat an Omni-generated ad as labeled AI content everywhere it goes, not just inside the paid campaign it was built for. A brand that wants some pieces of its content to read as unmistakably human-made, a founder story video, a behind-the-scenes clip, a customer testimonial, should keep that content out of the Omni pipeline entirely rather than generating it and hoping the distinction goes unnoticed.

Common Mistakes Advertisers Are Already Making

Treating the first generated storyboard as final. The whole point of the Refine with Prompts step is iteration. Google's own workflow description separates Generate Concepts from Refine with Prompts as two distinct steps precisely because the first pass is meant to be a starting draft, not a finished ad. Advertisers who publish the first storyboard Omni returns are skipping the step most likely to make the ad actually sound and feel like their specific brand rather than a competent but generic version of their category.

Ignoring the SynthID watermark implications for reuse. As covered above, an Omni-generated clip is not a blank slate you can quietly repurpose as unmarked organic content. Plan for where a generated ad is going to live for its entire lifecycle, not just its first placement, before you generate it.

Not checking performance against a human-made control before scaling spend. A generated video that looks polished in the Asset Studio preview is not the same thing as a video that is proven to convert for your specific audience. Run any new Omni-generated creative as one variant in a proper test, alongside an existing human-made or at least previously-validated asset, before committing meaningful budget to it exclusively. Google's own 1-Click Creative Testing feature inside Asset Studio, which launched alongside Omni, exists specifically to make this kind of control comparison low-friction, and skipping it in favor of just trusting the newest, shiniest asset is an easy way to quietly regress performance without noticing for weeks.

Anchoring against a weak or inconsistent brand page and accepting generic output anyway. Covered in detail in the Step 1 section above, but worth repeating as a mistake specifically: a rushed anchoring step produces safe, forgettable creative, and the fix (a cleaner reference page, a short written brand guide) takes minutes, not hours.

Assuming free means no rights or usage considerations. Omni being included in Google Ads at no additional cost does not automatically mean every generated asset is risk-free to use however you like across every channel and every market. Review your own legal and platform-specific requirements around AI-generated and AI-disclosed content the same way you would for any new creative production method, particularly if you operate in a regulated category or a market with specific AI-disclosure rules already in place.

Where This Fits Your Broader Creative Workflow

It is worth being precise about what Gemini Omni actually replaces and what it does not. It replaces, or at least meaningfully shrinks, the production gap between "I have no video budget" and "I have a video asset I can run in Google Ads." What it does not replace is a brand's need for a broader, consistent content pipeline that extends well beyond the inside of Google Ads itself: product photography for a growing catalog, organic YouTube content with its own thumbnails competing for clicks outside any paid auction, background music for social posts, and a consistent visual identity across owned channels that a Google-Ads-scoped tool was never built to manage.

This is the part of the workflow where a platform like Miraflow AI is a natural complement rather than a competitor to Google's own in-platform tools. Google's Asset Studio is, by design, scoped to what runs inside Google Ads campaigns. Miraflow's Cinematic AI Video Generator, built on Veo 3 and Veo 3.1, and its AI Image Generator, covering text-to-image, image-to-image editing, and inpainting, serve the same underlying need for fast, on-brand video and image creative, but for a brand's owned and organic content: social posts, YouTube uploads, product pages, and anywhere else a brand needs to look consistent outside a Google Ads auction. A brand generating a paid Demand Gen ad through Omni today still needs a matching product image for its Shopify listing, a thumbnail that gets an organic YouTube upload clicked, and a consistent visual voice across every channel that ad eventually drives traffic toward.

That is where Miraflow's YouTube Thumbnail Maker rounds out the organic side of the same video strategy, letting a brand generate a thumbnail from a prompt, optionally using their own product photo or face, right alongside whatever paid video creative Asset Studio produced for the same campaign push. And if that same product launch needs a short piece of background music for a social cut or a founder-voice video that should not run through an AI voice pipeline, Miraflow's AI Music Generator covers that piece without pulling a brand's whole content calendar back into a single-purpose ad tool. None of this claims Miraflow replaces Google's ad-platform-native generation, the two solve genuinely different jobs, one produces the ad that runs inside Google's auction, the other produces everything else a real content pipeline needs around it. Used together, a small brand can run Omni for its Google Ads creative while keeping its organic presence, product catalog, and broader brand look consistent through a single browser-based workflow at miraflow.ai rather than stitching together five separate tools for five separate asset types.

If you are also working through the other Performance Max and Google Ads AI changes that shipped around the same window, our guide to Performance Max's 2026 steering, reporting, and negative keyword controls covers the account-settings side of the same year's updates, and if a video an advertiser already has needs to become several vertical clips instead of one, Miraflow's AI Clipping tool turns a single long-form video into ranked, captioned shorts without needing a fresh generation for each cut.

What We Still Don't Know: Early Results Since the Broader Rollout

Because the August 26 rollout is recent, there is not yet a body of independently published, third-party case studies with real conversion or ROAS numbers specifically attributable to Gemini Omni-generated video versus an advertiser's previous creative. Coverage since the May preview and the summer rollout, including reporting from Search Engine Land and ppc.land, has focused on describing the workflow, the model, and the rollout timeline rather than reporting hands-on performance results from advertisers who have run it against a real control. That is worth being upfront about rather than inventing a lift percentage to make this post feel more conclusive than the current evidence supports.

What Google has published is a company-reported adoption statistic rather than a performance-lift number: based on Google's own internal data measured from January through June 2026, Google states that over half of small and medium-sized businesses advertising on Google Ads globally now use Google AI in some form to create or optimize their ad creative. That is a real, attributed number worth taking seriously as a signal of how quickly AI-assisted creative production has become the norm rather than the exception among smaller advertisers, but it describes adoption of AI creative tools broadly, not a verified, independently measured performance result specific to Gemini Omni video generation itself. Treat the adoption stat as a company-reported data point, and treat any specific conversion or ROAS claim you see about Omni until an independent source publishes one with a described methodology, which is the same standard worth applying to any single-source marketing statistic. Until then, the responsible framing is the same the 1-Click Creative Testing feature is built to encourage: test Omni-generated creative against your existing best-performing asset in your own account, and let your own data decide, rather than assuming a new tool with Google's backing automatically outperforms what you already have.

Frequently Asked Questions

Is Gemini Omni the same feature as the automatic video resizing Google added to Performance Max in August 2026? No, and this is the single most common point of confusion. The PMax resizing feature, which we covered separately, takes a video you already uploaded and generates missing aspect-ratio versions of that same video. Gemini Omni generates an entirely new video from a brief and your brand assets, and does not require an existing video to start from.

Does Gemini Omni cost extra inside Google Ads? No. Google has positioned it as a free feature rolling out globally in English inside Asset Studio, at no additional cost beyond your normal Google Ads spend.

Which campaign types can I deploy Omni-generated video into? Per Google's described workflow, finished assets can be packaged and deployed directly into Demand Gen, Performance Max, or YouTube campaigns from inside Asset Studio.

What happens if my brand does not have a strong, consistent landing page to anchor from? Anchoring quality directly affects output quality. A thin or inconsistent brand page tends to produce safe, generic-feeling video rather than something that reads as distinctly yours. Supplementing the URL import with a short written brand guideline document and two or three reference images meaningfully improves the anchor.

Can I edit an Omni-generated video after it is created? Yes, through the Refine with Prompts step, using natural language to adjust scenes, voiceover, pacing, and aspect ratio. It is less precise than a frame-by-frame timeline edit, so budget for a few rounds of prompted refinement rather than expecting the first pass to be final.

Does the SynthID watermark on generated videos cause any problems if I reuse the ad elsewhere? It is not a visible mark and does not degrade video quality, but it does mean the asset is identifiable as AI-generated wherever it goes, including if you repurpose it as organic content on another channel. Plan for that disclosure implication before reuse rather than after.

Should I trust a generated video over creative I already know performs well? Not automatically. Run any new Omni-generated asset as a tested variant against your existing best performer, using Asset Studio's 1-Click Creative Testing feature or your own experiment setup, before shifting meaningful budget toward it.

Conclusion

Gemini Omni closes a real, specific gap for advertisers who have a product and a brand but never had a realistic path to video creative: it generates a genuinely new, on-brand video ad from a brief rather than reshaping something you already made, which is exactly what separates it from Performance Max's automatic aspect-ratio resizing. The four-step workflow is straightforward on paper, but the quality of what comes out depends heavily on how well you anchor your brand in step one and how much you actually use the refine step rather than accepting the first storyboard. Treat it as a genuinely useful new production option, test it against what you already know works, mind where a SynthID-marked asset ends up being reused, and remember that a Google-Ads-native tool was never going to be the whole answer to a brand's broader content pipeline. For everything an ad campaign eventually drives traffic toward, the product pages, the organic YouTube presence, the social content, that same on-brand generation approach is worth having outside Google Ads too, and that is exactly the gap Miraflow AI's Cinematic AI Video Generator, AI Image Generator, and YouTube Thumbnail Maker are built to fill.