JD.com's Research Institute recently announced the open source of its self-developed JoyAI-Image-Edit image model, marking a new stage in AI photo editing technology that moves from traditional flat processing to 3D spatial modeling. As the first open-source model in the industry that emphasizes "spatial intelligence," it gives AI the ability to truly understand and reshape physical space.
Deep Modeling of 3D Space
The model closely aligns with the physical laws of the real world, comprehensively modeling dimensions such as camera perception and object displacement. This allows developers to directly call inference code to achieve precise spatial editing while maintaining geometric consistency of the scene.
JoyAI-Image-Edit has overcome long-standing challenges in spatial understanding in the open-source community, featuring high recognition. Its core highlight is the ability to flexibly adjust the camera's yaw angle, pitch angle, and zoom level based on natural language instructions.
Empowering Diverse Application Scenarios
In addition, the model supports continuous perspective movement, generating logically coherent roaming sequences. While maintaining overall structural stability, it can also scale or move specific objects, ensuring natural light shadow and occlusion relationships.
Aside from its groundbreaking spatial capabilities, the model fully supports 15 types of general editing needs, including object addition and removal, and style transfer. It has been widely applied in fields such as e-commerce production, creative design, and embodied intelligence, providing key underlying technical support for the industry.
