The LaVi-Bridge project is designed to seamlessly integrate diverse language models with generative visual models without the need for training, aiming to achieve text-to-image generation. LaVi-Bridge employs LoRA and adapters, offering a flexible plug-and-play approach that is compatible with a wide range of language and visual models. ConsiStory is a new text-to-image model that addresses the challenge of image consistency, generating coherent images without the need for training. Playground has released version 2.5, focusing on enhancing aesthetic quality and portrait details, outperforming other models in performance. Peking University, Stanford, and PikaLabs have jointly released a new open-source text-to-image framework, solving the challenges of text-to-image generation and surpassing the performance of SDXL and DALL·E3.