Google launched Gemini Omni on May 19, 2026 at I/O (source). It is a multimodal model that takes text, images, audio and video in, and produces video out — grounded in what Google calls "real-world knowledge." Physics, object permanence, and continuity hold up through long shots. Character consistency works across edits.
For most of the AI commentariat this was a Veo 3.1 / Sora-replacement story. For an Amazon brand owner running $200K–$2M/month, it is something more direct: the cost basis for PDP video, Brand Story video, and Sponsored Brands Video creative just dropped by roughly an order of magnitude, and Veo/Runway/Kling no longer have a moat as the "premium" option.
Here is what actually changes, what I would do this week if I were running an Amazon brand, and what I would ignore.
What I will not waste your time with
I am not going to walk you through the Gemini Omni feature set. Read the Google post or the 9to5Google writeup for that. What I want to talk about is the part that nobody on operator Twitter is talking about: the cost-and-time delta of Amazon-shelf-ready video, and what that does to your creative budget on a 90-day horizon.
What actually changed for an Amazon operator on Tuesday morning
Before May 19: a 30-second Brand Story video clip, decently shot with B-roll, animated product, and lifestyle inserts, ran $4,500–$12,000 through a creative agency, took 3–5 weeks, and was tested A/B against… usually nothing, because nobody wants to commission another $8K video to A/B against the first.
After May 19 — with Omni Flash already live in the Gemini app, Google Flow, and YouTube Shorts — you can generate the lifestyle and animated portions of that same video from a product photo and a written shot list, in roughly 45 minutes, for under $30 in compute. The output is not yet a 1:1 replacement for a real photoshoot for hero shots. But for B-roll, in-context scenes, "product in hand," and motion graphic backgrounds, it is already at parity or better.
I ran a quick test on Friday. Input: one clean white-background hero image of a kitchen tool we ship through one of my brands, plus a 6-line scene description. Output: 8 seconds of grounded lifestyle motion — a hand reaching across a butcher block countertop, picking up the product, using it. The physics on the product handle were correct. The lighting was consistent across the cut. Total cost: $0.42 in Omni Flash credits, ~6 minutes.
I am not saying this replaces the real photoshoot. I am saying it replaces the 15 B-roll seconds that the photoshoot was going to bill for at $400/clip.
The operator math
For a brand running 12 active SKUs with Brand Story video, Premium A+ video module, and Sponsored Brands Video creative, the typical annual video creative spend lands in the $80K–$180K range. Roughly the split:
- Hero product photography (still required, on-camera): 25%
- B-roll / lifestyle / in-use: 35%
- Motion graphics / text overlays / lower thirds: 15%
- Editing, color, sound: 25%
What Omni eats: chunks of the B-roll/lifestyle category and most of the motion graphics. Call it ~35–45% of that annual budget that is genuinely re-routable to AI generation inside 90 days. On a $120K annual creative budget, that is $42K–$54K moved from agency line-item to compute line-item — and the compute line-item is roughly $1,500/year for the same output volume.
The bigger lever is not cost. It is test velocity. Once you can generate a B-roll variant in 6 minutes instead of a 3-week reshoot, the marginal cost of A/B testing a Brand Story video — or a Sponsored Brands Video creative — collapses. Velocity Sellers' data on PDP video shows an 8–14% CVR lift from carousel video alone (placement matters more than length). Brands that have video have rarely tested it. With Omni in the workflow, you can run 4–6 video creative variants per SKU per quarter without renegotiating with a production house.
That is what changes. Not the cost line. The iteration loop.
Why most brand owners will read this wrong
The dumb take making the rounds: "AI can write a script and generate a video, replacing the creative agency entirely."
That is not what is happening. Gemini Omni is great at lifestyle motion, product-in-context, kinetic graphics, and consistent character animation. It is still inconsistent on:
- Brand-specific product detail (your logo, your specific colorway, exact label text)
- Hero-quality hero shots
- Tight 1080×1080 product crops where pixel-level fidelity matters
- Anything where the legal/compliance team needs to vouch for the claim ("supplement bottle showing FDA label")
The real signal is this: Omni did not replace the creative team. It replaced the production schedule. If you previously needed a 3-week production cycle to test a new Brand Story video, that cycle is now 3 days. The creative direction, the brief, the merchant decisions about what the customer needs to see — those still come from people who understand the shopper. The "shoot it" step is what got compressed.
The other dumb take to ignore: "Veo and Runway are now dead." They are not. They are different tools optimized for different jobs — Runway Gen-4.5 still has stronger frame-level editing control, and Kling 3.0 still wins on certain physical-realism categories. The shift is that there is now a third (very capable, Google-distribution-backed) horse in the race, which is going to drive Veo and Runway prices down 30–50% inside 6 months. That is the part to plan around.
What I would do this week if I ran a $200K/month Amazon brand
Five moves, in order:
1. Audit your current video assets by replaceability. Pull every active video — Brand Story, A+ video module, SBV creatives, PDP main carousel video. Tag each clip as: (a) hero product shot that needs to stay live-action, (b) B-roll/lifestyle that could be Omni-generated, (c) motion graphic / text overlay that could be Omni-generated. You will find 60–75% of your seconds are in category b or c.
2. Build one Omni-generated A/B variant per active SBV creative. Pick your top 3 Sponsored Brands Video creatives by ad spend. Generate one new variant for each that swaps the B-roll for Omni-generated lifestyle motion, keeping the hero product shots intact. Run for 21 days. Measure CTR and CVR delta against the original. Budget: ~$200 in compute and 6 hours of internal time.
3. Reframe your creative brief for Omni-first inputs. The brief that worked for an agency in 2025 ("we'll figure out the shots on set") is the wrong input for Omni. The brief that wins now is: hero product photo + 6–10 lines of explicit scene description per cut + tone reference. If your creative team or agency cannot deliver that brief format in 48 hours, they are about to fall behind.
4. Renegotiate your video production contract. If you are on a retainer with a video production house at $6K–$12K/month, that retainer was priced for a 2024 production cycle. Have the conversation now. The right structure is a smaller retainer for the live-action shoot blocks (one quarterly day with the camera, hero shots, key talent) plus an AI-assisted post-production layer. Anyone who tells you the retainer is fixed is going to lose your business in 9 months anyway — you might as well have the conversation first.
5. Add video to SKUs that did not have it. This is the underrated move. Most operators have video on hero SKUs and nothing on the long tail because the cost-per-SKU did not justify it. With Omni-assisted production, the cost-per-SKU for a basic 15-second PDP video drops to under $200 in compute plus 90 minutes of work. If you have 40 SKUs and 12 currently have video, the next 90 days you can have video on 30+ of them.
What I would ignore
- The "AI video is replacing humans" debate. Not relevant to the operator decision today.
- The Gemini Omni vs Veo 3.1 vs Runway Gen-4.5 vs Kling 3.0 benchmark race. Pick whichever your team can actually use inside an Amazon workflow. For most operators, that is whichever has the cleanest API and the lowest per-second compute cost — which today is Omni Flash, but that will shift.
- The "watermarking and provenance" feature debates. They matter for journalism and policy. For an Amazon Brand Story video, they do not change your shipping decision.
- The Sora 2 timeline. OpenAI shut Sora down on April 26, 2026 and is winding the API down by September. It is not coming back in time to matter for your Q3 creative plan.
The 90-day picture
Six months from now, the brands that did this work in May–July will have:
- A library of 40–80 Omni-generated B-roll clips, brand-consistent, tagged and reusable
- A creative brief workflow that goes from idea to live A/B test in 5–7 days, not 21
- 30–50% lower agency video spend, re-routed to test velocity and SKU coverage
- Video on long-tail SKUs that have never had it
The brands that wait will read this exact same article in November, wonder why their video-creative agency just raised retainers 15% to "absorb AI tooling costs," and start the conversation 6 months behind.
This is not the kind of AI news that asks you to add another tool to your stack. It is the kind that quietly resets the cost basis of an entire line-item on your P&L. Move while the gap is wide.