LumiVid targets HDR production

High dynamic range video stores more brightness information than standard video, which gives colorists more room to recover highlights and shape shadows in post-production. LumiVid is a new Lightricks research system that aims to generate that kind of footage from standard-dynamic-range input. (arxiv.org) The paper, “HDR Video Generation via Latent Alignment with Logarithmic Encoding,” was posted to arXiv on April 13, 2026 by authors from Lightricks, Gear Productions, and Tel Aviv University. It describes LumiVid as an adaptation framework built on a pretrained standard-dynamic-range video diffusion model. (arxiv.org) Instead of retraining a model from scratch for high dynamic range, the authors say LumiVid uses LogC3, a logarithmic camera encoding used in cinema workflows, to map high-dynamic-range imagery into a form the existing model already handles well. The paper says this avoids changing the model architecture or retraining its variational autoencoder. (arxiv.org) The output format matters to post-production teams. The researchers say LumiVid produces OpenEXR files, a format widely used for visual effects and grading pipelines because it preserves scene-referred image data instead of baking in a finished look. (arxiv.org) Video models often fail on consistency from frame to frame, which shows up as flicker, drifting textures, or changing faces. The LumiVid paper says its method targets native video generation rather than processing frames one by one, so it inherits temporal coherence from the underlying video backbone. (arxiv.org) Lightricks has been pushing its LTX line toward production workflows beyond short silent clips. Its current open-source documentation for LTX-2.3 says the model generates synchronized video and audio in a single model, with stable motion, consistent identity, and frame-to-frame coherence. (docs.ltx.video) The company’s public materials also frame LTX as a tool for “production-ready” video. The LTX-Video repository says LTX-2 can generate synchronized audio and video in one pass and can reach native 4K at up to 50 frames per second, while LTX Studio’s March 5, 2026 release notes say LTX-2.3 improved visual detail, audio quality, prompt understanding, and image-to-video consistency. (github.com) (ltx.studio) LogC3 is not a random technical choice. Arri’s workflow materials list LogC3 alongside LogC4 in color-grading guidance, which places LumiVid’s encoding choice inside an established camera and finishing pipeline rather than a format invented only for machine learning. (arri.com) The paper is still an arXiv preprint, and arXiv notes that material on the site is not peer reviewed. For now, LumiVid reads less like a shipping product launch than a sign that LTX’s video stack is being aimed at editors, colorists, and visual-effects teams that work in high-dynamic-range pipelines. (arxiv.org 1) (arxiv.org 2)