Scaling visual production using Amazon Bedrock's Stability AI Image Service

This post was written by Alex Gnibus of Stability AI.

The Stability AI Image Service is now available on Amazon Bedrock and offers ready-to-use media editing capabilities provided via the Amazon Bedrock API. These image editing tools extend the functionality of stable AI's stable diffusion 3.5 model (SD3.5) and stable image core and ultra model.

A professional creative production process consists of multiple editing steps to obtain the exact output you need. Amazon Bedrock's Stability AI Image Service allows you to modify, enhance, and convert existing images without jumping between multiple systems or sending files to external services. Everything is carried out through the same Amazon bedrock experience you already use. The impact on your business can be immediate for teams generating visual content at scale.

In this post, we explore examples of how these tools can accelerate professional-grade visual content with accurate creative control.

Editing tools available on Amazon Bedrock

Stability AI Image Services spans nine tools across two categories: editing and control. Each tool typically handles specific editing tasks that require specialized software or manual intervention.

Edit: Advanced features of granular editing procedures

Edit category tools make complex editing tasks more accessible and efficient.

The suite starts with a basic yet powerful retouching tool. For example, the Erase Object tool intelligently maintains background consistency while removing unnecessary elements from the image. The next animation introduces an erase object tool that removes mannequins from product shots while saving the background. This tool can convert the source image based on the mask image and derive the mask from the source image's alpha channel.

The Delete Background tool automatically separates subjects with accuracy. This allows you to create a clean, professional product list with consistent backgrounds and different lifestyle settings. This is an e-commerce game changer.

The following example shows the removal of the background of an image, but retains the furniture product details in the foreground.

Search, readjustment and exchange tools target specific elements in the image for changes. Search for object color changes and reorder. For example, view different colorways of your dress without a new photo shoot. The following image changes the color swatches for Search and Reklo furniture.

Search and exchange allow you to completely exchange objects. This will help you update seasonal elements in your marketing material and exchange products. Below is a search and exchange application for a virtual tryon experience.

The InPaint tool intelligently changes an image by filling or exchanging a specified area with new content based on the content of the mask image.

Control: Structural and Stylistic Accuracy

Tools in this category provide accurate manipulation of image structure and style through three special tools.

Sketch tools convert sketch style rendering into photorealistic concepts. Architecture companies can use this to transform conceptual drawings into realistic visualizations and turn apparel brands into design sketches into product mockups. This tool helps accelerate creative production processes, from early concepts to final visual execution.

In this example, the sketch tool transforms the building's architectural drawings to help real estate developers visualize concepts against the city landscape.

In another example, the sketch tool converts the drawing of a mannequin into a photorealistic model shot.

Structure tools maintain structural elements in the input image while allowing content changes. This tool changes subjects and styles while preserving layouts, compositions, and spatial relationships. Creative teams can use structural tools to recreate scenes across a variety of subjects and render new characters while maintaining consistent framing.

The following example illustrates a structural tool that transforms a workshop scene into a new scene, while maintaining its composition and spatial relationships.

Style Guides and Style Transfer Tools help your marketing team create new images tailored to your brand style and guidelines. The Style Guide tool takes artistic styles and colors from reference style images and generates new images based on text prompts.

In the following example, the Style Guide tool takes clues from the brand's color palette and textures and generates a new image to match the brand's identity.

The Style Transfer tool uses visual characteristics from the reference image to save the original composition while transforming existing images. For example, home decor retailers can change their product images from modern minimalists to traditional styles. Marketing teams can create seasonal variations by applying different visual styles to existing product catalogs.

Solution overview

To demonstrate stability AI image services on Amazon Bedrock, proceed with the example using the Jupyter notebook in the GitHub repository.

Prerequisites

To follow this, you must have the following prerequisites:

Create a Sagemaker AI notebook instance

Complete the following steps to create a Sagemaker AI notebook instance: This can be used to run the sample notebook.

1. Sage Maker AI Console, Navigation Pane, Under Applications and IDEschoose Notebook.
2. choose Create a notebook instance.
3. for Notebook instance nameenter the name of the notebook instance (for example, ai-images-notebook-instance).
4. for Notebook instance typechoose ml.t2.medium.
5. for Platform Identifierchoose Amazon Linux 2.
6. for The role of IAMselect an existing IAM role. AmazonSageMakerFullAccess and AmazonBedrockLimitedAccess Attached policy or select Create a new role.
7. Note the name of the IAM role you chose.
8. Leave other settings as defaults and select Create a notebook instance.

After a few minutes, SagemakerAI creates a notebook instance and its status changes from Pending In Inservice.

Ensure that the IAM role in the notebook instance has the required permissions

Complete the following steps to ensure that the Sagemaker AI Execution role you assigned to your Notebook instance has the correct permissions:

1. IAM console, navigation pane, under Access Managementchoose role.
2. in role Enter the name of the Sagemaker AI execution role that you used when creating the search bar, the notebook instance.
3. Please select the role of IAM.
4. under Permission Policycheck AWS managed policies AmazonSageMakerFullAccess and AmazonBedrockLimitedAccess It exists.
5. (Optional) If any of the policies are missing, please select Add permissionsselect Attach the policy Attach the missing policy.
   1. in Other Permissions Policy Enter the search bar, and the policy name.
   2. Select a policy and then select Add permissions.

Run the notebook

To run the notebook, complete the following steps:

1. Sage Maker AI Console, Navigation Pane, Under Applications and IDEschoose Notebook.
2. Please select the newly created one ai-images-notebook-instance Notebook instance.
3. Wait for the notebook to come in Inservice situation.
4. Select Open Jupyterlab A link to launch JupyterLab in a new browser tab.
5. In git Please select the menu Clone the repository.
6. Enter the URI https://github.com/aws-samples/stabilityai-sample-notebooks.git Select Include submodules and Download the repository.
7. choose clone.
8. In file Please select the menu Open from the path.
9. Please enter the following:
   stabilityai-sample-notebooks/stability-ai-image-services/stability-ai-image-services-sample-notebook.ipynb
10. choose open.
11. When prompted, select the kernel conda_python3select Select.
12. Go through each notebook cell and experience the stability AI image service with Amazon Bedrock.

cleaning

Stop to avoid ongoing fees ai-images-notebook-instance SagemakerAI notebook instance created in this walkthrough:

1. Sage Maker AI Console, Navigation Pane, Under Applications and IDEschoose Notebook.
2. Select ai-images-notebook-instance The SagemakerAI notebook instance you created.
3. choose actionselect Stop.

After a few minutes, the notebook instance will migrate from Stop In Stop situation.

4. choose actionafter that erase.

After a few seconds, Sagemaker AI will delete the notebook instance.

For more information, see Cleaning up your Amazon Sagemaker Notebook Instance resource.

Conclusion

Stability on Amazon Bedrock The availability of AI image services is an exciting step towards creating and manipulating visual content, and it has a time-saving impact, especially on the professional creative teams of companies.

For example, in media and entertainment, creators can quickly enhance scenes and create special effects, while marketing teams can easily generate variations of multiple campaigns. Retail and e-commerce companies can streamline product photography and digital cataloging creation, allowing game developers to prototype their environments more efficiently. Architecture companies can instantly visualize design concepts, while educational institutions can create more engaging visual content.

These tools allow businesses of different sizes to generate professional grade and engaging visual content with efficiency and creativity. These tools streamline operations, reduce costs, open up new creative possibilities, enable brands to tell stories more effectively and attract customers in a more persuasive way.

To get started, check out the stability AI models of Amazon Bedrock and AWS Sample Github Repo.

---

About the author

Alex Gnibus He is the Product Marketing Manager at Stability AI, connecting dots between cutting-edge research breakthroughs and real use cases. Experience from creative agencies to Deep Enterprise Tech brings both technical expertise and an understanding of the challenges professional creative teams can solve with generative AI.

Isha Doa I am a senior solutions architect based in the San Francisco Bay Area. She helps AWS Enterprise customers grow by understanding their goals and challenges and guiding how they can architect their applications in a cloud-based way, making sure they are resilient and scalable. She is passionate about machine learning technology and environmental sustainability.

Fabio Blanco He is Senior Customer Solutions Manager, Amazon Web Services (AWS) and Strategic Advisor, helping customers achieve business transformation, drive innovation through generative AI and data solutions, and navigate the cloud journey well. Prior to AWS, he was responsible for product management, engineering, consulting, and technology delivery roles across multiple Fortune 500 companies in the industry, including retail and consumer goods, oil and gas, financial services, insurance, aerospace and defense.

Sleman Patel He is a senior solution architect at Amazon Web Services (AWS), with a special focus on machine learning and modernization. With both business and technology expertise, Suleman helps customers design and build solutions that address real business problems. When he's not immersed in his work, Suleman loves to explore the outdoors, travel in the roads, and cook delicious food in the kitchen.