LightCompress provides off-the-shelf compression techniques for AI generated content models. It implements state-of-the-art algorithms including quantization, pruning, token merging, and token reduction to reduce model size and improve inference efficiency while maintaining model performance. The toolkit supports popular models like LLMs, VLMs, and diffusion-based video generators, with papers published at EMNLP 2024 and AAAI 2026.