in 3D rendering denoising engines like Optix and OIDN are able to generate and use Normal and Albedo channels (sometimes refered to as multilayers, AOVs.) This helps the denoising engine recognize details and preserve details/edges.
While this won’t apply to regular video footage it could make a huge difference for 3D renders if it could work with the Topaz model. Perhaps a toggle and either ability for Topaz to generate the channels natively or through input of user rendered maps/textures.
Video for quick reference: