Feed-forward single-image relighting model with physically based light control, using intrinsic conditioning from a coarse 3D scene.
Updated EU status checked Last shown
Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
PIXLRelight (also written PixlRelight) is a research method for controllable relighting of a single photograph. It was described by Miguel Farinha and Ronald Clark in an arXiv preprint (2605.18735, 2026). The user specifies a target light, such as a movable flash or a spotlight, and the model produces the photo as if lit by that light.
The method links physically based rendering with learned image synthesis through a shared intrinsic conditioning: albedo, diffuse shading and non-diffuse residuals. During training, this conditioning is computed from paired multi-illumination photographs. At inference, it is computed from a path-traced render of a coarse 3D reconstruction of the input under the user-specified PBR lights. A transformer-based neural renderer then applies the target illumination to the source photo, using per-pixel affine modulation to preserve fine detail. According to the authors, it runs in under a tenth of a second per image and reaches state-of-the-art relighting quality.
The project page offers an interactive demo with a cursor-controlled light and qualitative comparisons against prior work. A creator review states that code has been released, but no repository link appeared in the provided material. Weights, license and pricing details are not confirmed.
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
Single-photo relighting system where the user drags a light source or sets a spotlight. It estimates rough 3D scene geometry, relights it in Blender, and uses that as reference for the final prediction; code released.
Closest neighbours by meaning, not just by category.
Model
Research model for low-light image enhancement with continuous, controllable brightness, built on Flux 2 Klein.
Model
Research model for video relighting that jointly generates relit video and per-frame environment maps, no camera poses needed.
Research · Google
Research method for free-view rendering and relighting of full-body dynamic humans from four sparse RGB views.
Research
Research method that retargets human hand-object demonstrations to dexterous robot hands using one multi-reference RL policy.
Model · NVIDIA
NVIDIA's open-source image upscaler, v1.5, for latent, Flux and Z-Image outputs, targeting fast 2K results on consumer GPUs.
Model
Research model that unifies 3D scene generation and reconstruction in a single pixel-space diffusion model.
