In-Context LoRA (IC-LoRA) extends diffusion transformer models with an adaptation mechanism enabling zero-shot task generalization for visual content generation. The repository provides pretrained models, training configurations, and a ComfyUI workflow for tasks including film storyboarding, visual identity design, and visual effects. It also releases IDEA-Bench, a benchmark evaluating generative models on 100 real-world design tasks across 275 cases.