Personalized advertising with Mackevision’s AWS-powered render-on-demand system

The era of configurable automotive advertising

Mackevision, an Accenture Interactive company, is comprised of visualization specialists with a long history in developing vehicle configurators for the automotive industry. Traditionally, they’ve pre-produced CG content so it can be served from a web server. 

The assets in this system have to be pre-rendered using an offline renderer, which means the number of images you can provide with different configuration options is limited. At a certain scale, pre-producing all the content for truly personalized experiences is not economical. 

“On-demand rendering is our answer to that,” says Georg Wieland, Senior Manager, Global Production at Mackevision. “By extending the range of configuration options, we not only control the configuration of the car, but the complete staging of the assets: locations, perspectives, languages, time of day, weather…you can call it ‘experience configuration’”.

The result is a cloud-hosted, on-demand, real-time render service: an engine that can churn out marketing collateral in seconds, supercharged by the power of cloud computing. “The service uses AWS cloud infrastructure and Unreal Engine to produce configurable pictures and moving images on demand,” explains Wieland. “You no longer need to pre-produce millions of images: with on-demand rendering, you can get any content, any time, out of a single Unreal application.”

Use cases for this render service are virtually endless. Any application, such as a web frontend, can request configurable content. Content requests are simply sent to the render-on-demand system’s API gateway. If the requested image already exists in the cache Amazon Simple Storage Service (S3), it gets returned to the user right away. Otherwise, the request gets added to a render queue Amazon Simple Queue Service (SQS).​ The Unreal Engine application, running on AWS GPU instances with NVIDIA T4 cards, monitors the queue, renders the images, uploads them to the cache, and returns them to the user. 

In order to perform these requests, Mackevision built a plugin for Unreal Engine that handles all the communication with the AWS cloud infrastructure. This plugin is now available to the open-source software community, providing an easy way for creators to connect their Unreal Engine projects to the whole AWS infrastructure.  

The system features smart auto-scaling of render instances to deal with virtually infinite concurrent users. It also leverages Mackevision’s Configuration Backbone (CoBa) to manage complex configuration logic—not only for product configuration, but the entire experience. With this system, photorealistic high-resolution visualizations of elaborate experiences can be returned within milliseconds. 

In practical terms, it might look something like this: A car salesperson at a local dealership has some vehicles they’ve had hanging around for a while. They want to do a marketing push for those specific cars to shift them from the lot and into buyers’ driveways. 

They log on to a custom-made website that acts as the frontend for Mackevision’s render-on-demand system. Entering the identification number of the vehicle they want to sell brings up a CG replica of that car, which the salesperson can quickly configure so it exactly matches the spec of the one sitting outside their window. 

In just a few clicks, they can add different backgrounds, take shots from different camera angles, and create short video clips. The result is a marketing package of images and videos of the specific car they’re looking to sell that can be posted online, on social media, or used for ads. 

This made Unreal Engine’s ray tracing functionality a key tool in creating the solution. “Unreal’s ray tracing system helps us to produce highly photorealistic images,” says Wieland. “It is both powerful and flexible. Shadows and reflections have never looked so real in a game engine before. It also allows us to mix ray tracing with traditional game rendering features, achieving the performance and the quality we want.”

The second part of the puzzle is the ability to produce these world-class photorealistic images at lightning speed. Car configurators need to deliver images instantaneously. To achieve this, the team has put a lot of effort into performance optimization—not only within Unreal Engine, but also in the cloud infrastructure. This included optimizing request handling, logic resolution, caching, and data transfers.

Because Unreal Engine’s entire source code is available on GitHub, the team could integrate custom code and extend the application’s abilities to achieve its goals. Unreal Engine also provides two methods to create custom functionalities: C++ and Blueprint visual scripting. Both have been used extensively by Mackevision to create a highly performant application.
 

Personalization of digital assets 

The potential applications of Mackevision’s render-on-demand solution stretch far beyond automotive. The technology is agnostic to clients or industries, with the potential to transform marketing content creation across fashion, consumer goods, retail, industrial equipment, and many other sectors. 

With the pandemic shifting user behavior towards digital channels, many of Mackevision’s clients are transforming as well. 

As a society, we’re turning away from the 20th-century model of mass marketing and mass communication. That model was based on a combination of broadcast conversations and selective, limited listening.  In order to stand out today, it's critical that companies differentiate their brand experience and are relevant with the content they offer.

In short, we’re now in an era where personalization is key. Mackevision’s new render-on-demand system is in perfect sync with this new world, offering a way for companies to provide customized, relevant content at scale, in a way that’s both fast and cost-effective.