Car manufacturer Volkswagen Group has built a GenAI image content engine with Amazon Web Services (AWS) to power marketing assets across its brands, including Volkswagen, Porsche, and Ducati.
The GenAI venture was created in partnership with AWS Generative AI Innovation Centre – the firm’s $100 million programme to help its customers build and deploy custom GenAI solutions.
The Volkswagen solution will let the car company’s marketers generate realistic images of its vehicles at scale.
In a statement, AWS said that launching a single vehicle would typically need months of production work and costly six figure photoshoots. This type of image generation aims to take production from weeks down to minutes.
Brand Language Accuracy
AWS said that a challenge was making the assets from the company’s ten brands align with their individual brand identity and language.
The company identified two other challenges – the models lacked decades of Volkswagen design language, and the models had no knowledge of unreleased vehicles which limited their utility for forward-looking marketing campaigns.
The solution, AWS said, was fine-tuning foundational models on Volkswagen’s own visual data using AI data hub SolidMeta. This lets the content engine be fuelled by training data with control over vehicle specifications and environmental conditions.
AWS revealed the system uses a separate layer to break down the car images into its individual components, including its wheels, grills, headlights and mirrors. It then scores the different parts of the cars against reference images to ensure 100 percent accuracy.
The company also used brand guidelines to develop synthetic prompts that were both compliant and non-compliant, so that the model could distinguish between the two.
This example is one of many where brands are using GenAI to produce realistic marketing assets, driving both cost and time savings, making an arduous manual process into a scalable marketing engine.
