August 21st.SenseTimeAnnounced todayOpen SourceLightweight, original uniformMultimodal large model Sensenova U1.5 Lite。
According to Sensenova U1.5 Lite native support 3-4k context length, capable of addressing multiple constraints, such as subject matter, number, spatial relationship, text, layout, style, etc., at the same time, to enhance the stability of the implementation of complex visual tasks。
In addition, Sensenova U1.5 Lite has the following characteristics:
- Better-quality visual generation• Improvement of the overall picture, colour, material, light, real sense and local detail and reduction of “locally correct but not fully complete”。
- More reliable raw image editing• Enhancement of the main identity, spatial structure, layout relationship and non-editing area, while increasing the capacity of local modifications, replacement of elements, precision of text and editing of multiple reference maps。
- Better Text and Complex Layout• Strengthening of the capacity of Chinese and English languages, posters, information maps, brand visualization and multilingual layouts, from “production of content” to “organization's full visual expression”。
- More sophisticated visual control: Supporting more accurate editing of designated areas and objects, such as Bounding Box, Visa Marker and single and multi-graph references。
- NATIVE 4K HIGH RESOLUTION OUTPUT: Take into account, in high-resolution generation, the whole image and the very fine muscle, tiny text and light reflection。
Officially, as 8B ARGUMENTSThe light-quantitative model, SenseNova U1.5 Lite, exceeded the equivalent model in command compliance and image editing maintenance, and reached the delivery level of comparable large-scale business models in text rendering and complex layouts。
1AI WITH OPEN SOURCE LINK:
- GitHub: https://github.com/OpenSensenova/SenseNova-U1
- Hugging Face: https://huggingface.co/collections/sensenova/sensenova-u15
- Modified Community: https://modelscope.cn/models/SenseNova/SenseNova-U1.5-8B-Mot