CGI animated films and VFX relie on massively detailed 3D scenes featuring hundreds of lights, high resolution textures and many complex shaders. Pre-visualization of such assets involve complex computations, usually requiring the use of production renderers such as MentalRay or RenderMan. However, the non-interactive render times (up to severalhours per frame) preclude any interactive setup of scene contents and lighting. Graphics hardware (GPU) accelerated solutions exists, they are generally based on progressive ray-tracing methods and thus very accurate to render shadows and reflections. However, they generally require the author to model the scenes in a specific way, for instance using proprietary shaders or lights, in order to get proper pre-viz. In this way they are not compatible with a standard production workflow based on Maya3/MentalRay. Other proprietary solutions exists but they are generally tightly bound to large CGI companies workflows and hardly applicable to other pipelines. In this talk, we present Technicolor’s GPU pre-viz system compatible with standard Maya/MentalRay production workflows. This GPU-based pre-viz system has been developed upon Technicolor’s research work in the field of real time rendering and pre-viz technology.