Google's Gemini Omni Brings AI Video to Asset Studio: What Indian D2C Brands Should Do

What Google actually shipped on Aug 25

For years the honest answer to "why aren't you testing more video?" was money and time. A single product film meant a shoot, an editor, and a week of back-and-forth before it ever saw a campaign. On Aug 25, 2026, Google put a real dent in that excuse.

Google announced that Gemini Omni, its multimodal model from DeepMind, now lives directly inside Asset Studio in Google Ads. In Google's words, you can "generate, edit and scale on-brand video directly in Google Ads." No export to a separate tool, no stitching creative back into your campaigns by hand. The video is born where your ads already live.

This isn't the same thing as the automatic video resizing Google rolled out earlier in the year, which took an existing clip and reshaped it. Gemini Omni makes the clip in the first place, from a brief and your brand assets, then lets you rework it in plain language.

The four steps, from brief to campaign

The workflow Google describes has four moves, and it's worth knowing them before you touch the tool:

First, brand anchoring. You import your brand guidelines and point it at your website URL so the model has your colours, tone and product context.

Second, concept generation. From a text prompt or a static image you already own, it builds video storyboards and motion scenes. So a single festive banner can become the seed for a moving ad.

Third, refinement. This is the useful part. You adjust scenes, voiceover, pacing and aspect ratio using natural language rather than rebuilding creative from scratch. "Make the opening two seconds faster, swap the voiceover to a calmer read, hold on the product shot longer" is the kind of instruction it takes.

Fourth, deployment. It exports multi-format video, including horizontal 16:9 and vertical 9:16 versions tailored to Google and YouTube surfaces, and you push it into your campaigns.

Where it plugs in: Demand Gen, PMax and YouTube

Google says Gemini Omni works with Demand Gen, Performance Max and other Google and YouTube campaigns. For a D2C brand that matters, because your video probably has to earn its keep across prospecting on YouTube and Demand Gen and then inside the PMax asset group that's doing your bottom-funnel work. One brief, several cuts, all of it landing where you already spend.

Every generated clip carries SynthID, Google's imperceptible watermark that flags the content as AI-made. That doesn't change how the ad performs, but it's worth knowing the provenance is baked in.

Why this matters more for D2C than for big brands

A large brand with a retainer agency and a content calendar was never really blocked by video cost. A ten-person D2C team selling skincare or kitchenware on Shopify absolutely was. That's why this lands harder for you.

Google shared one number worth sitting with: over half of small and mid-sized businesses on Google Ads globally now use Google AI to create or optimise their creative, based on its own data from January to June 2026. Whatever you think of AI video, your competitors in the feed are already leaning on it.

The video bottleneck most Indian D2C teams hit

Here's the pattern we see across Indian D2C accounts. The brand has three or four hero videos, all made months ago. They run until they fatigue. CPMs creep up, hook rate falls, and there's no fresh cut ready because the next shoot is "planned for after Diwali." The account stalls on stale creative, not on bidding.

Gemini Omni attacks that specific bottleneck. If you can turn a product still into five hook variations in an afternoon, creative fatigue stops being a quarterly crisis and becomes something you top up weekly. The value isn't that the AI video is more beautiful than your shoot, it usually won't be. The value is volume and speed of testing.

Put a rough number on it. A single edited product video from an agency in a metro can run ₹15,000 to ₹40,000 and take a week. At that price you test one idea a month, maybe. If the same still image seeds five usable hook cuts in an afternoon at effectively media-only cost, your testing budget suddenly buys learning instead of production. For a brand doing ₹5–10 lakh a month on Google Ads, that shift from one test a month to one a week is often the difference between an account that plateaus and one that keeps finding new winners. If you want a second opinion on whether your account is actually creative-limited or something else, that's the kind of thing our performance marketing team pressure-tests before touching budgets.

Gemini Omni vs your current video workflow

Before you replace anything, be clear about what each path is good for. AI generation and a real shoot are not enemies, they solve different jobs.

Read that table as a division of labour. Keep your shoot budget for the one or two anchor films that carry your brand story and your real customer voices. Point Gemini Omni at the churn: the endless hook tests, the festive refreshes, the format variants you never had time to cut.

The rollout plan for your account

Don't open the tool and start generating twenty clips on day one. Treat it like any new lever: small, measured, and judged against a control.

Week one: brand anchoring done right

Garbage in, garbage out applies hard here. Spend real effort on the brand-anchoring step. Upload your actual brand guidelines, not a rushed one-pager, and point it at a product page that represents your best work. If your guidelines don't specify tone, do-not-use claims, and logo rules, write them first. The model can only respect boundaries you've actually stated.

Week two: test against your human-made control

Pick one PMax asset group or one Demand Gen campaign. Keep your best human-made video running as the control. Add three Gemini Omni variations built from different hooks, same offer. Give it enough budget and time to exit the learning phase before you read anything into the numbers, and watch hook rate and view-through, not just ROAS on day two.

Week three onward: scale only what beats the control

If a generated clip beats your control on hold and cost per result, scale it and generate three new challengers from the winning angle. If nothing beats the control, you've learned your control is strong, which is also useful. This is ordinary creative testing discipline, the AI just makes the "generate three new challengers" step cheap.

Where AI video still gets D2C wrong

The honest caveats matter, because this is where founders get burned. Even Google's guidance says you should scrutinise whether AI-generated videos genuinely meet your brand standards rather than trusting the model's interpretation of your guidelines. That's not marketing modesty, it's a real limit.

COD, regional language and trust cues

Indian D2C carries baggage that generic AI video doesn't understand. A COD-heavy audience reads trust cues that a model trained on global creative won't prioritise: the returns promise, the "pay when it arrives" reassurance, the human face that makes a first-time buyer comfortable. AI voiceover in Hindi or a regional language can land in an uncanny middle ground that quietly kills conversion even when the visuals look fine.

So keep a human in the loop on anything that touches price, delivery promises, or a specific claim about results. Use the tool to make ten hook openings, absolutely. Don't let it write the closing line that promises "delivery in 2 days" without someone checking that's true for the pincodes you actually serve. And route any generated clip that mentions an offer through the same approval your WhatsApp broadcast copy goes through, so your channels tell one consistent story.

What to do this week

Check whether Gemini Omni has reached your Asset Studio yet, since a global rollout arrives account by account. If it has, write or clean up your brand guidelines before you generate a single frame. Then pick one campaign, one control video, and three AI challengers, and run a proper test instead of a vibe check.

The teams that get value here won't be the ones generating the most video. They'll be the ones who treat AI video as a testing engine bolted onto disciplined creative measurement, keep their shoot budget for the films only humans can make, and never let a generated clip make a promise to a customer that a person hasn't verified.