Stable Diffusion 3.5 hits Amazon Bedrock: What it means for enterprise AI workflows

Creating beautiful generative AI imagery is fun and useful, but it’s not all businesses need.

Corporate text-to-image generation is more than just creating images. It involves integration with existing workflows and other enterprise AI tools. this is a direction Stable artificial intelligenceThe vendors behind Steady Proliferation understand this.

Today, Stability AI and Amazon Web Services (AWS) jointly announced: Stable diffusion 3.5 large Available on Amazon Bedrock service. AWS is the only public cloud service offering a flagship stable AI model.

This move isn’t just for simple usability. It’s about integration and go-to-market strategies that inject more energy into Stability AI’s efforts as the company’s new CEO Refocus on meeting customer needs. Amazon Bedrock provides a single, unified API that allows enterprises to access and use multiple AI models, including stable diffusion models. This is important because AWS’s own research shows that most enterprises use multiple models simultaneously. Users including the NFL and Stride Learning have already benefited from this approach.

Stable Diffusion 3.5 As it deploys Amazon Bedrock, Stability AI faces challenges with Google, Midjourney, Ideogram and Flux Pro by Black Forest Labswait. The company aims to stand out through greater image diversity in terms of style, timely compliance and corporate workflow.

“We chose AWS for a reason,” Stability AI CEO Prem Akkaraju said in an exclusive interview with VentureBeat. “Because that’s where developers and creators are, and we want to bring our tools and models to where they are. place. Our goal is to support professional content creators.

AWS has its own image generation technology, why does it need Stable Diffusion?

The increasingly competitive field of text-to-image generation artificial intelligence also includes models from Amazon. Early December, Amazon Rising Star Artificial Intelligence A family of models has been released, including image generation models.

Baskar Sridharan, vice president of artificial intelligence and machine learning services and infrastructure at AWS, told VentureBeat that multiple text-to-image generation models provide users with choices. Amazon Bedrock provides consumers with a single unified API so they can choose to use the same API to deploy any model available on the platform. Sridharan also pointed out that the model evaluation tools provided by AWS can help enterprises choose the best tool for a specific deployment.

Not surprisingly, Akkaraju considers the Stable Diffusion 3.5 to be superior to other models. Stability AI has backed up this claim with reported timely compliance benchmarks.

“Our research shows that Stable Diffusion 3.5 Large leads the market in timely compliance, allowing models to closely follow a given text prompt, truly making it a top choice for efficiency and high-quality performance,” said Akkaraju.

How enterprises can integrate Stable Diffusion into Amazon Bedrock AI workflows

Since the end of October, Stable Diffusion 3.5 Large has been available to Stability AI users through the company’s API and its Stable Assistant service.

Stability AI is not aware of any real price differences between using Stable Diffusion on its API or using Stable Diffusion through Amazon Bedrock. For enterprise users now using Stable Diffusion on Amazon Bedrock, the real benefit is the ability to adapt to larger, more complex enterprise workflows. Enterprises can benefit from a unified workflow that connects multiple models from different vendors with a single API.

The NFL is already using this approach.

The NFL has an app called “My Cause, My Cleats” that uses Amazon Bedrock to enable a community-focused collaborative experience to create custom sneaker designs. Sridharan explained that the app uses both Anthropic Claude and Stable Diffusion. The NFL uses Claude to create detailed prompts to understand user preferences and determine what they want. This prompt is then fed into a stable diffusion to produce the image. The entire process and workflow is done on Amazon Bedrock, eliminating the need to jump between different services.

Another organization benefiting from consolidation are education providers Learning in strides. The company needed images to power its online learning game Legends Library—a massive amount of images, up to 1,000 images per minute. Amazon Bedrock provides the infrastructure to support operations at scale. In addition to high-performance scale, there is also a need to ensure the safety of the image-generating output. That is Amazon Bedrock Guardrail API Sridharan noted that with guardrails in place, Stride Learning is able to meet responsible image-generating AI policies.

“When you do it all using a single API endpoint, customers can build these applications very easily,” Sridharan said.

An eventful 2024 for Stability AI and the road ahead with new leadership

Stable Diffusion 3.5 is updated and launched on Amazon Bedrock, capping an extraordinary year for Stability AI.

Founder and former CEO Emad Mostak resigns in March amid concerns about a lack of focus and revenue. It wasn’t until June that Stability AI found a permanent replacement, appointing Akkaraju.

Akkaraju has been responsible for a series of model updates so far. He also helped bring in new investors like Napster founder Sean Parker and advisors including prominent directors James Cameron September. Akkaraju has a background in visual effects and has helped produce films including Cameron Avatar. In his view, the professional visual media industry will completely transform in the coming years, moving towards generating content rather than rendering content.

“We do a lot of work in the creative industries, which is a good thing because these are probably the most demanding clients we can ask of,” Akalaju said.

Looking to the future, he joked that Stability AI plans to take over the world. More seriously, though, he hopes the company will continue to innovate and work toward meeting real workflow needs.

“We will continue to move this model forward,” Akalaju said. “You might even see us release the next image model again next year and stay on top of the trend.”