BluePixel enables fast precise 4K image editing with an intelligent agent
BluePixel: Toward Native 4K Agentic Image Editing
Computer Vision and Pattern Recognition
Summary
Editing ultra-high-definition 4K images is hard because a single tool struggles with detailed control and vague user commands. The authors present BluePixel, which combines a powerful 4K editing model with a smart agent that breaks down instructions and verifies results. This approach makes training affordable and editing faster, completing complex edits in under 9 seconds while preserving image quality. BluePixel performs very well in tests compared to other 4K editors, especially in keeping parts of the image unchanged when they shouldn’t be edited.
What this means in practice
- •For digital content creators: Perform fast, high-quality 4K image edits guided by natural language instructions with better control over non-edited areas.
- •For interactive media developers: Integrate controllable and efficient native 4K image editing into applications using automated decomposing and verification of user commands.
Authors
Shanyan Guan, Shiyu Li, Xin Li, Zhizhou Chen, Yunzhe Xu, Yuehao Liu, Ruixue Shen, Shiju Zhang, Min Jin, Qiang Liu, Junbo Liu, Penggang Shi, Sheng Sun, Xiao Wang, Yong Zhang, Zhongang Qi, Wei Li, Chao Ma, Ying Tai, Yanhao Ge
Abstract
Instruction-driven image editing has reached native 4K, yet how such editors are trained remains undisclosed, and pretraining one from scratch is beyond most teams. The task itself compounds the difficulty: a single end-to-end editor cannot make precise parametric adjustments, match specialized experts, or interpret the vague instructions users actually write. We argue that 4K editing is a system problem rather than a single-model problem, and present BluePixel, which couples a native-4K editing model with an agent harness; the complete training recipe is disclosed in this paper. The model makes 4K training practical: it turns a 1K-pretrained editor into a native-4K one through progressive VAE adaptation that keeps 4K training as affordable as 1K, absorbs a long tail of vertical editing skills through continual learning, and refines the aesthetic-fidelity trade-off with reinforcement learning. Distillation via source-anchored target projection (SATP) then compresses 50 denoising steps to 4, completing a 4K edit in 8.7 s on a single Kunlunxin P800 XPU while preserving task success rate and background consistency. The harness, BluePixel Agent, makes editing controllable: it decomposes user instructions into specialist and general operations, verifies every result with the same reward model that guided training, and retries within bounded attempts. On VINS-4KEval++ (3,305 native-4K cases), BluePixel ranks second among seven models, behind only GPT Image 2 in non-edited-region consistency.