SenseTime launches open-source lightweight multimodal model with 4k output capabilities

SenseTime has officially announced the open-sourcing of SenseNova U1.5Lite, a new lightweight, native unified multimodal large model. The model, with 8 billion parameters, showcases strong performance in various core visual tasks, surpassing models of similar scale and achieving results comparable to large commercial models in complex layout and text rendering scenarios. The model natively supports context lengths of up to 4K, enabling it to manage multiple constraints such as subject, quantity, spatial relationships, text, layout, and style simultaneously. This significantly enhances stability during the execution of complex visual tasks. SenseNova U1.5Lite introduces several key upgrades. It offers improved visual generation, enhancing overall composition, color, material, lighting, realism, and local details. This addresses common issues where traditional models produce locally correct but overall incomplete results. Additionally, the model provides more reliable native image editing capabilities, improving the preservation of subject identity, spatial structure, layout relationships, and non-editable areas. It also enhances the overall performance of local modifications, element replacement, text refinement, and multi-reference image editing. The model excels in handling text and complex layouts, strengthening its capabilities in Chinese and English text, posters, infographics, brand visuals, and multi-text formatting. This pushes content generation toward complete visual expression. In terms of visual control, it supports methods such as Bounding Box, Visual Marker, and single or multi-image references, allowing for more precise editing of specified regions and objects. Notably, it achieves native 4K high-resolution output, maintaining both the overall macro composition and extremely fine textures, small texts, and lighting effects in high-resolution generation. The project is available at https://github.com/OpenSenseNova/SenseNova-U1.