This post was written by An Tran, Director of Engineering at Creative Fabrica overseeing product development, platform engineering and AI research.
Creative Fabrica is a digital marketplace that offers various design assets such as fonts, graphics, and templates. The core user base consists of crafters, graphic designers, artists, web designers, and anyone else who needs design assets for their projects. With over 4 million registered users from all over the world, the platform caters to both professionals and beginners who require high-quality design elements.
The Generative AI Opportunity
The field of generative AI began to rapidly accelerate around the release of Stable Diffusion (a deep learning model that can generate realistic images given a raw input text), bringing opportunities but also business challenges. The future supply of products at the core of Creative Fabrica’s marketplace was quickly shifting from human-scale content production (days and weeks by professional designers) to one that could conceivably happen orders of magnitude faster at machine-scale (seconds through generative AI models). This would effectively commoditize pieces of individual content and result in AI-generated content drowning out designer-generated content in sheer volume. The challenge here would be matching designers in terms of quality.
To better serve customers, Creative Fabrica decided to innovate and to take advantage of the recent progress of AI-powered content generation. The goal was to provide customer with even more complete content creation experiences, that are real time and adaptive to their specific, transactional needs. Creative Fabrica could effectively and seamlessly fill whatever gaps existed in the product catalog on the fly through generating results close to what customers are searching for rather than relying on matches with pre-built content.
Product Development
As Creative Fabrica’s plans around incorporating generative AI into their offerings began to take shape, there were essentially two things to be done to get a running start: First, quickly put basic image generation tools in front of customers to iterate and learn about what works with the community. Second, leverage external partners to build internal muscles for implementing and deploying AI tuned for Creative Fabrica’s audience.
Working with ML6, an AI consultancy and AWS partner based in Ghent, Creative Fabrica rapidly spun up Stable Diffusion as an internal service which allowed them to present to customers an experience that let them easily compare output from generative models of interest: Stable Diffusion, Midjourney and DALL-E 2. From there, Creative Fabrica was able to iterate by offering additional tools representing fine-tuned results for different categories of content familiar to users within Creative Fabrica’s Spark family: patterns, clip art and sketches. They were also able to offer the ability to create variations of existing images via the ImageMix tool and text generation from simple prompts for article writers.
Speed of Innovation with Amazon Web Services (AWS)
As Creative Fabrica worked to connect their user community to the fun and power of generative AI, they made full use of the capabilities offered by Amazon Web Services. Using AWS allowed them to immediately get access to all the necessary resources to quickly deploy out-of-the-box models as services and weave them into their customer-facing applications.
From a technical perspective. Amazon EC2 gave access to high-power GPUs as quickly as they could develop and deploy services. Amazon ECS automatically scaled the solution to meet spikes in user activity. Being able to use a combination of spot and on-demand instances allowed Creative Fabrica to be cost effective while we experimenting with deploying unique content pipelines. Lastly, Amazon Simple Queue Service enabled them to easily decouple the AI microservices from the rest of the technology stack. With the elasticity and scalability of Amazon ECS and Amazon SQS, Creative Fabrica could focus on the value and the speed of innovation, without having to worry about the overhead caused by having to manage a complex infrastructure.
Creative Fabrica leveraged AWS to go from concept to live features in the span of a few weeks. This allowed their customers to get their first experience of text-to-image generative AI, creating and sharing their creations with each other and forming a community on the platform around using these new capabilities to meet their creative needs.
Since the launch of Creative Fabrica Spark, customers have issued close to 10 million prompts, generating more than 40 million images, over 4 million of which they have published for sharing with others on the platform.
This post was written by An Tran, Director of Engineering at Creative Fabrica overseeing product development, platform engineering and AI research.