Omi: Scaling Brand Visuals with 3D Digital Twins

In a recent NVIDIA feature I had the opportunity to explain how Omi is redefining the way brands produce product visuals. As the co-founder and CEO of Omi, I talked with NVIDIA about how our virtual photo studio uses advanced 3D digital twin technology and NVIDIA GPUs to deliver high-quality, on-brand content at scale. In this article I’ll lay out the problem we set out to solve, the technology and partnerships that make it possible, how our workflow works, and what this means for brands that need visuals every day across multiple channels.

🎯 The daily problem: Brands need visuals every day, everywhere

Brands today face relentless demand for product visuals. Between e-commerce listings, social media, paid ads, email campaigns, marketplaces, digital catalogs, and packaging mockups, a single product can require dozens or even hundreds of images and videos in different formats, resolutions, and styles.

I often say the core challenge in modern retail and marketing is not ideation—brands know what they want to show—but logistics. Traditional photo shoots are expensive, slow, and inflexible. You book studio time, fly in products, hire talent, settle lighting and styling, and then post-process images. If the campaign changes, or if a new SKU is added, you may need to redo the shoot.

The result is a mismatch between speed and quality. Marketing teams need to move fast, but they also need images that are perfect and on-brand. Demand spikes around product launches, promotions, seasonal campaigns, and marketplace deadlines make it nearly impossible to consistently deliver top-tier product visuals using legacy approaches.

We built Omi to be a solution for that gap. Our goal was to let brands get a high-quality, on-brand visual for any product in minutes—without the logistical overhead of a traditional photoshoot. To do that at scale, you have to rethink the entire content production pipeline. That’s where 3D and cloud GPU compute come in.

🧩 What 3D digital twins are — and why they matter

When I talk about 3D at Omi, I’m referring to digital twins: precise, photo-realistic 3D replicas of physical products. A digital twin is not a stylized illustration. It’s a faithful virtual copy of the product with the same scale, materials, textures, decals, and finishes.

Why is that important? Because when a product behaves, reflects light, or shows texture digitally in the same way it would in real life, you can create studio-grade imagery without the physical product present. That unlocks several key advantages:

Scale: A single digital twin can be re-shot in any environment, angle, or lighting condition indefinitely.
Consistency: Every asset generated from the digital twin uses the same underlying geometry and materials, so brand appearance remains consistent across channels.
Speed: Once the twin exists, generating new images or variations is a software task rather than a re-shoot.
Cost-efficiency: You avoid repeated studio bookings, styling, and logistics for each variation.
Versioning and localization: Swapping artwork, packaging language, or finishes can be done digitally and at scale.

In practical terms, a digital twin is created from a combination of CAD models, photogrammetry captures, and texture work. We stitch high-resolution texture maps, calibrate materials with physical-based rendering (PBR) parameters, and rigorously match color and gloss to the real product. That means we pay attention to albedo (base color), roughness, metallicity, normal maps for surface detail, and subsurface scattering where appropriate.

Creating a convincing digital twin is not trivial. The modelling and texturing work demands expertise and compute. But once it’s done, the payoffs are enormous: a single twin powers thousands of image permutations, animations, lifestyle mockups, and interactive 3D viewers.

⚡ How we scale rendering with NVIDIA GPUs

High-fidelity 3D rendering is computationally expensive. If someone tries to run a full production-quality render on a consumer laptop, they can expect hours of processing per image. That’s incompatible with the fast turnaround times our customers require.

That’s why we turned to NVIDIA GPUs and cloud infrastructure. Our partnership gives us access to enormous, on-demand GPU capacity. To be concrete: we can scale up to thousands of GPUs—peaks of up to 4,000 GPUs—when a large batch of renders needs to be processed.

"We get up to 4,000 GPUs on demand with NVIDIA. No one else has this computing capacity on the market." — Hugo Borensztein, Co-founder & CEO, Omi

Why does that matter? GPUs accelerate ray tracing, path tracing, and the complex math involved in physically based rendering. Modern renderers take advantage of GPU parallelism to compute light transport, reflections, soft shadows, and global illumination far faster than CPUs. With the right GPU fleet, a production-grade image that would have taken tens of minutes or hours can be completed in minutes—or in some cases, seconds.

We run our rendering workloads on GPUs provisioned within AWS infrastructure, enabled by NVIDIA hardware. The ability to spin up large pools of GPUs on demand means we can treat any workload—whether a single image or tens of thousands—with the same operational model: scale to meet demand and return results quickly.

Over time, we’ve seen performance improvements in raw speed and cost-efficiency. Two years ago, a high-quality product photo might take 10 minutes to render. Today, thanks to advances in GPU hardware and renderer optimizations, many images render in seconds or a couple of minutes. That speed difference transforms how brands plan campaigns: the lead time for visuals shrinks dramatically.

🛠️ Inside the Omi virtual photo studio workflow

Let me walk you through how a brand gets from product to images using Omi. Our workflow is designed to be intuitive, integrated, and automation-friendly.

1. Capture or ingest the product information

The first step is creating the digital twin. That can happen in several ways:

Photogrammetry captures (multi-angle photos that are processed into 3D geometry and textures).
CAD or native 3D files provided by the product design team.
Manual 3D modelling when necessary for stylized or hard-to-scan items.

We often combine sources: CAD for correct dimensions and photogrammetry for accurate surface detail. We maintain rigorous color pipelines so the twin accurately represents paint, fabric, metallic finishes, and printed artwork.

2. Material calibration and PBR setup

Next, we calibrate materials with physically based rendering parameters. That involves assigning maps for albedo, roughness, metallic, normal detail, and any emissive or subsurface properties. We also set up decal layers for labels and artwork so we can swap or update branding without remodeling the product.

3. Presets, templates, and on-brand libraries

To be truly productive, brands need templates and presets. We build lighting setups, camera presets, and style guides that reflect a brand’s identity—color palettes, composition rules, shadow softness, and shot types (hero, packshot, hero with shadow, 45-degree view, top-down, lifestyle mockup, etc.).

Once presets are defined, marketing teams can request new assets using these templates and be confident the output will be on-brand. This reduces review cycles and ensures consistency across campaigns.

4. Cloud rendering and queueing

When a user requests a set of images, the render jobs are queued and dispatched to GPU instances. We scale the GPU pool based on demand. Large batch jobs—like product catalogs containing thousands of SKUs—can run in parallel because of the cloud GPU elasticity we access through NVIDIA on AWS.

5. Post-processing, color management, and delivery

After rendering, images undergo color checks, any required retouching, and automated cropping or resizing to meet channel-specific requirements. Assets are then pushed to the brand’s DAM (Digital Asset Management) system, e-commerce platform, or social scheduler.

6. Continuous iteration

If the creative brief changes, we don’t reshoot. We update the digital twin or swap artwork and re-render. That iteration is fast and inexpensive compared to physical production. For example, swapping a package label for a region-specific language or testing a seasonal colorway happens in minutes.

⏱️ Speed and the real-world impact on go-to-market

Speed is not only an operational metric; it changes business outcomes. When teams can get high-quality content quickly, they can run more experiments, optimize creative assets for conversion, and respond to marketplace events in near real-time.

Imagine a product launch approaching and creative direction changes at the last minute. With traditional photography, you may miss the start of a campaign. With Omi’s virtual studio, the same brand visual can be produced in minutes. I’ve seen brands shift from reactive to proactive: testing multiple hero images, A/B testing backgrounds, or generating regionalized imagery with local language and compliance details—all within the same day.

We’ve observed measurable improvements in time-to-market and campaign velocity for customers who switch to a 3D-driven approach. The ability to generate an asset "the day before you need to launch" eliminates many stressors and reduces expensive expedited shoots.

🤝 Our partnership with NVIDIA and the Inception program

Our collaboration with NVIDIA has been instrumental. As a startup, joining the NVIDIA Inception program provided not only access to technology but also a gateway to an ecosystem of innovation. NVIDIA’s focus on high-performance compute and their ongoing investment in developer tools and libraries gave us leverage that would have been difficult to achieve alone.

"Omi has been thrilled to be part of the Inception program for almost a year. For us, being able to partner with such a forward-thinking and innovation-driven company is just very exciting and makes a total difference." — Hugo Borensztein

NVIDIA’s GPU platforms, when combined with cloud infrastructure, deliver a few critical advantages:

Elasticity: We can provision thousands of GPUs on demand to handle peak workloads.
Performance: Advanced GPUs accelerate ray tracing and ML-based denoisers that reduce render times dramatically.
Software ecosystem: NVIDIA’s software stack, including SDKs and optimized renderers, allows us to squeeze more throughput and maintain predictable performance.
Support and community: Access to technical resources and a community of partners helps us iterate faster on product features.

It’s important to note that the hardware sits within cloud providers like AWS, but the difference in latency, throughput, and sheer compute capacity comes from NVIDIA’s GPUs. Without them, our model—high-quality, near-instant product visuals at scale—would be much harder to deliver.

💡 Real-world benefits and examples for brands

To make all this tangible, here are concrete scenarios where Omi’s approach produces clear value:

1. E-commerce marketplaces

Marketplaces have strict imaging guidelines for hero shots, white backgrounds, and multiple angles. With Omi, brands can deliver compliant images at scale without staging a single studio shoot. Any SKU update or packaging tweak can be re-rendered quickly to meet marketplace deadlines.

2. Social and performance advertising

Performance marketers need dozens of creative variants to test headlines, formats, and visuals. Digital twins let teams generate many variations—different backgrounds, copy overlays, and product colorways—rapidly. Faster iteration improves conversion rates and reduces media waste.

3. Retail and omnichannel catalogs

Retailers require consistent product imagery across print and digital catalogs. The Omi workflow ensures the same lighting and composition rules apply across channels, which helps maintain brand integrity while supporting mass distribution.

4. Localization and compliance

When you need to localize packaging or include region-specific regulatory text, doing it digitally saves time and cost. Each localized version can be rendered and validated without rebuilding the physical package.

5. Seasonal drops and limited editions

Limited editions often require distinct photography. Digital twins make it possible to visualize variants—different materials, engravings, or label artwork—without producing prototypes. That speeds decision-making and reduces physical waste.

🔍 Quality, consistency, and the automation challenge

Automating creative production introduces a tension: how do you preserve the artisanal look and brand nuance when the process becomes mechanical? We approach this problem with a combination of art and engineering.

We build style guides into the system—lighting rigs, camera perspectives, mood presets, and retouching rules that uphold the brand’s DNA. At the same time, we enable creative control: designers and art directors can tweak lighting angles, shadows, and material properties. The balance of automation (to scale) and manual controls (for taste and nuance) is essential.

When I describe the result, I like to emphasize fidelity. The digital twin needs to mimic the physical product so closely that a consumer cannot tell the difference between a real photo and our render. That requires rigorous color management and attention to micro-details like stitching on textiles, gloss variation, and edge reflections.

We also embed QA steps to catch deviations. Automated checks analyze render outputs for color accuracy, exposure, shadow fidelity, and alignment with brand rules. If something is off, it’s routed back to a human operator for review. This hybrid automation model lets us be fast while maintaining quality control.

💰 Economics: costs, scalability, and return on investment

One of the big questions I’m asked is: does this save money? The short answer is yes—but with nuance. The economics depend on volume, asset type, and how much you value speed and flexibility.

Here’s how I think about ROI for Omi customers:

Upfront costs: Creating a digital twin has a one-time cost. That cost varies by product complexity—simple items are cheaper, complex items with multiple materials or moving parts are more expensive to model and texture.
Per-asset cost: Once the twin exists, the marginal cost of producing each new image is largely compute and delivery. Cloud GPU rendering increases costs relative to local rendering, but the speed and scale benefits offset that when you need many assets quickly.
Operational savings: Brands save on studio time, photography teams, travel, and logistics. These savings become significant at scale.
Opportunity cost: Faster creative cycles mean shortened time-to-market and the opportunity to run more tests. That can increase revenue through optimized conversion and reduced media spend waste.

For many of our customers, the break-even point comes quickly—often within the first catalog or seasonal campaign—because the alternative is repeated photoshoots and the associated operational overhead. When you factor in the value of speed and the ability to run more creative experiments, the financial case gets stronger.

🚀 Technical considerations and challenges we overcame

Operating a production-grade virtual studio at scale involves multiple technical challenges. Here are a few we addressed and the approaches we used:

Rendering fidelity vs. speed

Higher fidelity means longer render times. We invest in render optimization techniques like smart sampling, denoising algorithms (often GPU-accelerated), and adaptive light sampling to preserve quality while reducing compute. Pre-baked lighting for static elements and importance sampling for critical regions also help.

Asset management and versioning

Managing thousands of twins and their variants requires robust metadata and versioning. We track every material change, label revision, and render preset so we can reproduce or roll back to any prior state. This is critical for auditing and legal compliance in regulated industries.

Interoperability and formats

We support a broad set of 3D formats (GLTF, USD/USDZ, OBJ, FBX) and ensure materials translate accurately across renderers. USD (Universal Scene Description) has emerged as a powerful option for complex scenes and efficient interchange between tools.

Color management and calibration

Color fidelity is paramount. We maintain color profiles and use spectrophotometer-based references during twin creation. We also standardize our color pipelines to ensure outputs match brand color standards across devices and print processes.

Security and IP protection

Brands are understandably sensitive about their IP—product designs, packaging, and artwork. We implement strict access controls, encrypted storage, and contract-level protections to safeguard assets and ensure only authorized users can request renders or export 3D data.

Scaling compute reliably

Managing thousands of GPUs across cloud providers requires orchestrating instances, handling transient failures, and optimizing for cost. We built robust queueing, retry logic, and cost-aware scheduling so that large batches are handled predictably and economically.

📈 Measurable outcomes and performance improvements

Across our customer base, we’ve seen clear outcomes that justify the transition to digital twin-driven production:

Render time reduction: Average render times for high-quality hero shots have fallen from ~10 minutes to under 2 minutes in many cases due to GPU improvements and renderer optimizations.
Faster campaign launches: Brands can produce and approve comms materials days—or sometimes weeks—faster than with traditional photography.
Higher throughput: Catalogs that once required weeks of shoots can now be generated in days.
Lower per-asset operational cost: Savings on studio logistics and retakes add up significantly across hundreds or thousands of SKUs.
Increased creative experimentation: Teams are running more A/B tests and creative variants, improving conversion metrics and lowering paid acquisition costs.

🔮 The future: where product visuals are headed

Looking ahead, I’m excited about a few trends that will shape the next phase of product visuals:

Real-time interactive experiences

Real-time rendering powered by GPU advancements opens the door to interactive shoppers who can rotate, customize, and preview products in near real-time. That interactive experience narrows the gap between online and in-store shopping.

Integration with AR and metaverse platforms

Digital twins are naturally portable to AR and other immersive formats. Brands can reuse the same digital assets across product pages, AR try-ons, and virtual worlds with minimal rework.

AI-assisted asset generation

Machine learning models are already accelerating tasks like background replacement, automated retouching, and even generating plausible materials. Combining ML with physically accurate renders will make content pipelines smarter and faster.

Standardization and industry ecosystems

Standards like USD and GLTF, combined with cloud rendering services, will create ecosystems where brands, agencies, and retailers can exchange high-fidelity assets seamlessly. That standardization reduces friction and supports interoperability.

Sustainability

Digital production reduces the carbon footprint associated with travel, shipping, and physical shoots. While cloud compute uses energy, consolidating production into optimized data center workflows can be more sustainable than repeated physical logistics.

📣 Final thoughts and call to action

At Omi, we’ve built a virtual photo studio because we believe brands shouldn’t have to choose between speed and quality. The combination of digital twins and scalable NVIDIA GPUs allows us to offer on-brand, high-fidelity product visuals in minutes rather than days or weeks.

Our work is a practical response to a real-world challenge: marketing teams need visuals every day, for every channel, and at every scale. By digitizing products and leveraging modern GPU-accelerated rendering, we remove the bottlenecks of traditional photo production and give teams the freedom to iterate more, launch faster, and keep brand integrity intact.

If you’re a brand wrestling with asset backlogs, inconsistent imagery, or slow campaign turnarounds, I’d invite you to consider whether a virtual studio model could be the right solution. Reach out, and we can discuss how a digital twin strategy might change your content operations and creative velocity.

Finally, I want to acknowledge the role NVIDIA and the Inception program have played in making this possible. For startups like ours, access to cutting-edge GPU technology and a supportive ecosystem has been a difference-maker. The ability to tap into vast GPU capacity—up to thousands of GPUs on demand—lets us deliver on the promise of fast, scalable, and stunning product visuals.

Thank you for reading. If you’d like to explore how Omi works technically or see case studies, I’m happy to share deeper details and walk through examples tailored to your industry.