GitHub radar
The SenseTime model can create and modify 4K images
SenseTime announced SenseNova U1.5, a universal model that creates and edits 4K images with text prompts such as adding/removing things and transferring styles from reference images.
“SenseNova U1.5 is a native unified multimodal checkpoint for more accurate, consistent, reliable and aesthetically pleasing visual creation.” SenseNova U1.5 is a multifunctional image generation and editing model which accepts a text prompt and outputs 4k-resolution images, or accepts a reference image and a text prompt and edits it according to the given instructions, such as adding or removing objects from the image, changing the style of the image or the location of the objects in the image based on complicated spatial relationships between objects. It also supports multi-reference image editing which enables seamless transfer of visual features and characteristics from multiple references into the target image. “It’s released under the Apache 2.0 license by SenseTime, a top AI research institute in China.”
Why a vibe-coder should care
This tool can help those who generate content like blog posts, ads, product images by giving you the ability to generate and edit images locally without paying any cloud or api usage fees. You can use the multi-reference feature to make sure that the entire series has the same character or brand looks, by telling the model what should be preserved in the image.
How to install
Copy this and send it to your agent — Claude Code, Codex, any of them:
Install SenseNova U1.5 from SenseTime on my machine: https://huggingface.co/sensenova/SenseNova-U1.5-8B-MoT — set it up for my hardware, run it, and show me how to generate an image from a text description and how to edit an existing photo with a text instruction.
A Mac with 32 GB of RAM or more, or a GPU with 24 GB or more — around 11 GB in 4-bit mode (estimated from parameter count).
Open on Hugging Face▌ More finds