The data to be translated: LangSplat is an innovative 3D language Gaussian model developed by researchers from Tsinghua University and Harvard University. This model achieves efficient and accurate open-language search in 3D space, being 199 times faster than the previous LERF method. Researchers have successfully captured object boundaries by visualizing learning features, while also demonstrating higher accuracy in tests. LangSplat is not only fast but also more accurately labels various parts and components of objects, such as the various ingredients in a bowl of ramen soup. In tests, LangSplat has shown superior speed and accuracy on both the LERF dataset and the 3D OVS dataset, bringing a new breakthrough to the field of 3D language search.