AI Image Editing Always Distorts? JD Open-Sources JoyAI-Image-Edit Model to Break Through 3D Spatial Editing Challenges

When using AI for image creation or editing, issues like object distortion when moving, incorrect perspective when changing angles, and occlusion chaos when adjusting object relationships are common—all due to AI's inability to understand the 3D spatial structure of images. Recently, JD Exploration Research Institute open-sourced its self-developed JoyAI-Image-Edit model, which integrates 'spatial intelligence' into AI image editing for the first time, solving the spatial logic problems of traditional models.

The model has three core spatial editing capabilities: perspective transformation (generating new views by specifying camera angles), spatial roaming (continuous view movement to produce coherent sequences), and object spatial relation control (moving objects while maintaining scene structure stability). It also supports 15 general editing functions and can be widely used in e-commerce content production, creative design, 3D model reconstruction, etc., enhancing the practicality of AI image technology.

This open-source release is a key step in JD's AI multimodal layout. Recently, JD also open-sourced JoyAI-LLM Flash, launched the OpenClaw series (with 455% weekly growth in token usage), and plans to build the world's largest embodied data collection center to promote deep integration of AI and industry.