AWS Serverless Generative AI: Amazon Nova Reel Foundation Model with Bedrock and Lambda

On December 3rd, 2024, during the re:Invent 2024 keynote, AWS announced support for five additional models designed for Generative AI use cases across text, image, vision, and video. With these new additions, alongside Amazon Titan, there are now 13 Amazon models available via the Amazon Bedrock service to address diverse business use cases. This count doesn't even include the other third-party models Bedrock supports, such as Anthropic, Stable AI, Meta, and many others!

The five new models announced today are:

• Amazon Nova Micro
• Amazon Nova Lite
• Amazon Nova Pro
• Amazon Nova Canvas
• Amazon Nova Reel

Excited about these advancements, I decided to update one of my transcript summarization APIs with the Nova Micro foundation model. This API leverages Amazon Bedrock and Lambda to deliver Generative AI capabilities. When I updated the underlying foundation model to Amazon Nova Micro, the entire process—updating, building, validating, and redeploying the API—took less than 5 minutes. Below is a link to an article describing this process:

https://dev.to/bhatiagirish/aws-serverless-and-generative-ai-with-lambda-and-bedrock-5fnl

In this article, I will demonstrate another model from the Amazon Nova family: Amazon Nova Reel. I will use the Nova Reel model to create a video clip from a static image, which will be provided as input to the model.

Let's follow the steps for this Generative AI use case using Amazon Bedrock, Lambda and Amazon Nova Reel foundation model.

Request Access to Amazon Nova Reel

First, you need to request the access to the desired model, in my case, the process took less than 20 seconds to request the access for Amazon Nova Reel model.

Amazon Bedrock Playgrounds

I used Amazon Bedrock Playground to experience with Nova Reel foundation model.

• From Amazon Management Console, navigate to Bedrock playground and select Nova Reel Model.
• You will need to accept/confirm for s3 as bucket will be created to store the generated video clip.

Input Image for the foundation model

Earlier this year, I took a NYC skyline camera shot. I will feed this camera pic to Amazon Nova Reel model and will ask to create a video clip.

Upload NYC Skyline Image

Upload your favorite image of NYC Skyline. This image will used by the model to generate the video clip.

Prompt engineering is an important aspect of Generative AI. Ensure your prompt is clear and precise to guide the model effectively. Below is an example prompt:

• "First person view aerial shot: NYC Skyline:4k;Cinematic;in a sunny day; dolly in"

Monitor the Generation Progress

The Amazon Nova Reel playground provides real-time progress updates as it generates the requested video clip.

Once the video clip is successfully generated, you’ll see an option to download it.

This video clip is also automatically stored in your S3 bucket. You can delete it from there so that you don't incur ongoing cloud cost for this s3 bucket.

Final Result (Static Image and Generated Video Clip)

This was another example of Generative AI use case. AWS continue to add models and it is exciting to see many options to pick from when it comes to foundation models!

In the coming days, I plan to explore the other Nova models announced today. Stay tuned for updates!

Thanks for reading!

Click here to watch the YouTube video for this solution:

https://www.youtube.com/watch?v=cLc4ZRrBZZg

𝒢𝒾𝓇𝒾𝓈𝒽 ℬ𝒽𝒶𝓉𝒾𝒶
𝘈𝘞𝘚 𝘊𝘦𝘳𝘵𝘪𝘧𝘪𝘦𝘥 𝘚𝘰𝘭𝘶𝘵𝘪𝘰𝘯 𝘈𝘳𝘤𝘩𝘪𝘵𝘦𝘤𝘵 & 𝘋𝘦𝘷𝘦𝘭𝘰𝘱𝘦𝘳 𝘈𝘴𝘴𝘰𝘤𝘪𝘢𝘵𝘦
𝘊𝘭𝘰𝘶𝘥 𝘛𝘦𝘤𝘩𝘯𝘰𝘭𝘰𝘨𝘺 𝘌𝘯𝘵𝘩𝘶𝘴𝘪𝘢𝘴𝘵