The research team at Tsinghua University has developed the LCM model, which achieves a 5-10 times speedup in text-to-image generation, breaking the limitations of traditional generative models. LCM supports image generation tasks under given conditions, significantly improving the efficiency of mainstream text-to-image models through potential consistency techniques. LCM-LoRA further expands LCM's sampling capabilities, providing solutions for models with different artistic styles. The open-source LCM code has been adopted by multiple platforms, with over 200,000 downloads on Hugging Face and over a million views, marking the entry of generative models into the real-time era.