Three-dimensional representation of real scenes has taken a qualitative leap with the arrival of Gaussian splats, a technique that models space using elastic primitives that capture color, density, and position. However, one of the most complex challenges in this field is separating what is intrinsic to an object's material (its albedo) from what depends on lighting (shading). This intrinsic decomposition, common in 2D image editing, now extends to 3D, allowing textures to be modified without altering the scene's light. Instead of copying previous approaches, researchers have proposed modeling decomposition as independent sets of Gaussian primitives, each adapted to the characteristics of its layer —albedo, shading, and view-dependent residuals— and guiding optimization with data-driven predictions to disentangle multiview photographs. The result is a workflow where the user can edit the texture of a flat surface simply by modifying the albedo in an image, and the intrinsic radiance field propagates the change to all angles with plausible lighting.
This innovation has direct implications for industries such as film, video games, architecture, and virtual reality, where the ability to alter materials without re-lighting the entire scene saves hours of manual work. From a business perspective, integrating these capabilities into custom application platforms makes it possible to offer advanced 3D editing tools that were previously only available to large studios. At Q2BSTUDIO, as a software and technology development company, we tackle projects where AI for businesses is combined with three-dimensional representation models to create disruptive solutions. Our team uses artificial intelligence techniques and AI agents to automate segmentation and editing processes, while the infrastructure relies on AWS and Azure cloud services to process large volumes of multiview data. Additionally, cybersecurity is key when handling sensitive digital assets, and business intelligence through Power BI helps measure the performance of these solutions. We achieve all of this thanks to custom software development that adapts to the specific needs of each client, whether it is a post-production studio or an augmented reality company.
Intrinsic decomposition based on Gaussian splats opens the door to applications that go beyond editing: from simulating dynamic lighting in virtual environments to the digital restoration of historical heritage. By being able to separate intrinsic color from light, re-lighting with synthetic sources is also facilitated, a field where business intelligence services can model predictive analysis scenarios. At Q2BSTUDIO, we believe that the convergence between computer graphics and artificial intelligence is unstoppable, and that is why we offer custom application solutions that integrate these algorithms efficiently and scalably. If you are looking to transform your 3D workflow or incorporate advanced editing and rendering capabilities, our team is ready to design the technological architecture that makes it possible.
