Lightricks has developed the IC-LoRA technology, which allows precise changes to lighting in exterior video footage using simple visual prompts.

What happened
Developers at Lightricks released IC-LoRA — a specialized module for the base video model LTX-2.3-22B. The technology allows controlling the direction and hardness of sunlight in the video. Control is achieved through a "light ball" that the user places in the corner of the frame, using it as a visual signal. The process is implemented as a single-stage solution, working in one step without the need to use heavy external models for post-processing each individual frame.
Context
Traditional methods of changing lighting in generative video often require either full regeneration of the content or the use of complex chains of multiple neural networks to process each frame, which leads to a loss of temporal consistency and an increase in computational costs. IC-LoRA offers a transition to a model of modular control of parameters through lightweight adapters.
Why this matters for the industry
For the video production industry, this means a transition from "black box" generation to the use of precise and controllable tools. The single-stage approach is critically important for maintaining temporal consistency of the video, as it eliminates artifacts that arise when processing frames sequentially with different models. This allows for the implementation of controlled lighting changes in professional production pipelines with minimal latency and without significant infrastructure complexity.
Why this matters for users
Users get an intuitive way to control generative video. Instead of trying to compose complex text prompts to describe the light, you can simply add a graphic primitive to the corner of the screen. This significantly simplifies the artistic correction process: you can make the sunlight softer or change its direction with a simple visual action, making the tool accessible to both professional editors and enthusiasts.
What is still unknown / limitations
No direct technical disagreements in the evaluation of the method have been identified, however, discussions in the community are shifting from the mechanics of operation to questions of using the method in production pipelines and potential legal risks.
Sources
Author
Look at AI, editorial team
