Generate Images and Video with vLLM-Omni on SageMaker

This guide shows how to deploy two SageMaker AI endpoints from a single AWS vLLM-Omni Deep Learning Container to generate images and video from text prompts. A real-time endpoint runs FLUX.2-klein-4B for image generation, returning a base64-encoded PNG inline. An asynchronous endpoint runs Wan2.1-VACE-1.3B, taking the generated image and a motion prompt to produce an MP4 stored in Amazon S3. The workflow supports both a command-line interface and an optional Streamlit application. Separate endpoints allow each model to use the instance type and inference option suited to its workload.