HubSpot Scales Image Generation 150% with Stability AI in Amazon Bedrock
Opportunity | Using Stability AI Models in Amazon Bedrock to Accelerate Creative Production at Scale for HubSpot
HubSpot helps businesses in more than 135 countries grow by connecting marketing, sales, and other customer-facing teams on a unified customer platform. Most customers use multiple Hubs across content, marketing, sales, operations, service, and commerce.
From high-performing infrastructure to intuitive application development tools, AWS provides a comprehensive set of AI capabilities to help businesses innovate with generative AI. HubSpot’s AI tools—collectively called Breeze—run on AWS infrastructure. With Breeze features in Content Hub, users can generate on-brand images directly within their content creation workflows.
HubSpot recognized that Stability AI’s Stable Diffusion 3 Large (SD3 Large) model could deliver professional-quality images at scale. “We were looking for a way to scale to meet our customers’ desire to generate images in high amounts,” says Sean Carroll, senior product manager at HubSpot. “We set an artificial cap of 120,000 images generated per month to make sure our infrastructure could handle the demand.”
The Stability AI team suggested accessing Stable Diffusion models through Amazon Bedrock to meet increased demand without technical complexity. “When you get to an enterprise level, scaling through an API can be a challenge,” says Sanwal Yousaf, solutions engineer at Stability AI. “But using Amazon Bedrock, it becomes really simple for companies to adapt and scale as needed.”
Interested in seeing how you can deploy generative image models like Hubspot?
Solution | Managing 150% Growth in Image Generation
SD3.5 Large can create photorealistic pictures, follow a wide range of artistic styles, or design fantastical imagery.
Using Stability AI models in Amazon Bedrock means that HubSpot can access the latest models through automatic updates; for example, automatically updating Stability AI SD3 Large to Stable Diffusion 3.5 Large (SD3.5 Large). Because Amazon Bedrock simply and securely integrates multiple generative AI models and automatically deploys the appropriate option for specific use cases, HubSpot has been able to help marketers transform their daily workflows.
Instead of spending hours searching through stock photo libraries or waiting weeks for professional shoots, customers describe what they need using natural language prompts. For example, a marketer writing a blog post can ask Breeze Content Agent to provide “a drone shot of a mountainous valley with clouds.” In the background, SD3.5 Large instantly produces an on-brand image for the post. The model can create photorealistic pictures, follow a wide range of artistic styles, or design fantastical imagery.
“Stability AI’s image generation models in Amazon Bedrock handle the nuanced requirements that make the difference between amateur and enterprise-grade results,” says Ryan Ellis, senior vice president and head of product at Stability AI. “We built our models with professional use cases in mind. Teams can test dozens of visual concepts, personalize content for different audiences, and respond to market opportunities faster.”
Customers can generate images within any Hub, although most visuals complement projects in Marketing Hub or Content Hub. For example, a user can ask a Breeze Agent to create a podcast and then generate an accompanying image with specific size parameters. The agent generates the podcast content through Anthropic’s Claude in Amazon Bedrock and the image using Stability AI’s Stable Image Core, an enhanced version of Stable Diffusion XL (SDXL)—both accessed behind the scenes in Amazon Bedrock while the user carries on with higher-level tasks.
“We want to use the best model for the right feature at the right time, as opposed to a one-size-fits-all approach,” says Carroll. “That level of flexibility and scalability of Amazon Bedrock is a value-add for us.”
Within 4 months, HubSpot customers generated 300,000 images, a 150 percent increase above the initial monthly cap of 120,000 images. But the impact extended far beyond image volume. In the year following the Stable Diffusion integration, HubSpot experienced record growth in its attach rate between the Content and Marketing Hubs—meaning that customers who used one Hub increasingly purchased the other.
“We want to use the best model for the right feature at the right time, as opposed to a one-size-fits-all approach”
Outcome | Driving Marketing Innovation for Customers Through Automatic Access to Cutting-Edge AI Models
Through its use of Amazon Bedrock, HubSpot stays ahead of the AI curve. When Stability AI and other leading AI companies like Anthropic release new models, HubSpot customers can access them instantly. This makes it possible for HubSpot to continue providing innovative new features, such as image-editing tools or sketch-to-image, which transforms simple sketches into detailed pictures.
“Using Amazon Bedrock, we always get the latest and greatest model,” says Carroll. “It’s very simple for us to stay on top of any new developments in AI.”
Getting Started
For organizations interested in utilizing Stable Diffusion to power their businesses, please contact us to inquire about an Enterprise License. You can also visit our solutions page on AWS Amazon Bedrock to start building scalable solutions.
Our flexible licensing options allow you to deploy our models in your environment with advanced customization and full control over your data.
