Generate images and video with vLLM-Omni on SageMaker AI – Part 2
aws ·
This article demonstrates how to generate images and animate them into videos using vLLM-Omni on Amazon SageMaker AI. The workflow deploys two endpoints from a single AWS Deep Learning Container: a real-time endpoint for the FLUX.2-klein-4B image model and an asynchronous endpoint for the Wan2.1-VACE-1.3B video model. Users provide a text prompt to create a still image, which is then passed along with a motion prompt to generate a short MP4 video. The solution utilizes Amazon S3 to handle the larger multipart payloads required for video generation via asynchronous inference. A command-line interface and a Streamlit application are provided to facilitate this end-to-end process. The authors highlight that separating image and video tasks allows each model to use the optimal instance type and inference pattern for its specific latency and payload requirements.
aws