Accessing the Bedrock Console Part 2

In this lesson we continue exploring the Amazon Bedrock console, focusing on discovering and testing image and video generation models from the Model Catalog and using the Playground to run quick experiments. We’ll cover model details you should watch for (such as parameter counts and token limits), how to run image and video generation jobs in Playground, and what to expect for region-based availability. We’ll start by filtering the Model Catalog to only show image-generation models.

​

Inspecting an image model (Stable Diffusion)

I opened the Stable Diffusion entry because it’s a widely used family for text-to-image generation. The model overview highlights useful technical details you should note when designing prompts:

• Model size: 8 billion parameters (for this entry).
• Hosting: serverless (no provisioning required).
• Input limits: about 77 tokens — typical of CLIP-style encoders (see CLIP for details).

These details help with prompt design and understanding model capabilities and limits.

​

Generating images in Playground

From the Model Catalog you can open Playground to run quick, interactive tests. Example workflow:

1. Choose the provider (e.g., Stability AI) and the model (e.g., Stable Diffusion).
2. Pick a profile (for example, On-Demand) and click Apply.
3. Enter a descriptive prompt — e.g., “Solar system showing Earth’s position relative to the Sun.”
4. Generate the image, then inspect, create variations, or download the resulting asset.

Playground makes it easy to iterate on prompts and quickly evaluate model outputs.

​

Video models and region availability

Video-capable models are less common and often depend on the selected AWS region. In the Oregon region (us-west-2 at the time of this lesson) there was only one video model listed: Ray by Luma AI. Note that Ray may be deprecated in that region — model lifecycle and availability can change. If a given video model is deprecated or unavailable, switching regions is a common workaround. For example, switching the console to US East (N. Virginia) surfaces Amazon’s Nova Reel, which supports text-to-video and image-to-video workflows. Clicking the Nova Reel entry shows lifecycle information (this model was marked legacy when captured). “Legacy” means it may be retired later, but it can still be functional for testing while available.

​

Generating video in Playground (Nova Reel example)

Steps to generate video in Playground:

1. Open Playground and select the Nova Reel model (choose Output: Video and the desired version, e.g., 1.1).
2. Apply the model. Playground will warn that video generation runs as a background job and that the output will be stored in an S3 bucket.
   • If you don’t have an S3 bucket set up, the console can prompt you to create one.
   • Note: S3 storage incurs standard S3 charges in addition to Bedrock request charges.
3. Enter a prompt (for example, “sports car driving through a city at night”) and submit the job.
4. Video jobs can take a few minutes. Playground shows the job history and the S3 bucket location for the resulting asset.

In my test the job completed in roughly two minutes and produced a short clip that matched the prompt.

​

Model availability summary

Model availability varies significantly by region and provider. Use the Model Catalog to verify which modalities are offered in your selected region and check the model lifecycle state (stable, legacy, deprecated) before relying on a model for production.

​

Model access and permissions

The console used to include a “Model access” page where you requested access to third-party models. That page has been retired: most serverless models are now enabled by default, and access control is handled through IAM policies. Use IAM to control which users and roles can call specific models.

​

What this enables for your team

• Faster developer onboarding to Amazon Bedrock without needing to write code first.
• Rapid, interactive experimentation with different foundation models via Playground.
• Faster prompt engineering and identification of candidate models for your use case.
• Reduced development risk by validating model behavior in a sandbox before integrating into production.

Key takeaway: the Bedrock console (Model Catalog + Playground) provides a safe, interactive way for teams to discover, test, and iterate on foundation models available in a region so they can identify the best models and prompts for their applications. This concludes the short lesson on using the Bedrock console for model discovery and quick experimentation with text, image, and video modalities. If you want to automate workflows or integrate Bedrock into CI/CD, the next step is to access Bedrock programmatically via the command line or SDKs. Further reading and references:

• Amazon Bedrock documentation: https://docs.aws.amazon.com/bedrock/
• Amazon S3: https://aws.amazon.com/s3/
• AWS Identity and Access Management (IAM): https://aws.amazon.com/iam/
• Stability AI (Stable Diffusion): https://stability.ai/
• Luma AI (Ray): https://www.luma.ai/