SenseTime open-sources new image-text model

SenseTime has open-sourced a new model called U1-8B-MoT-Interleaved. It's designed to create picture books, storybooks, and multi-page slideshows without losing character consistency or art style. The model improves how images match text descriptions and reduces visual flaws. It can also automatically generate multi-page PPTs by pulling out key points and handling layout and text. The model is now available on Hugging Face.