Generating Video Highlights Using the SmolVLM2 Model