Hi, why does the MRQ (movie render queue) always render everything until the end of the render range (instead of starting at start-frame)? I know it could be useful for non-deterministic effect that could be randomized or deterministic one that required procedural simulation as there are no bake/cache system.
But I work in a company that make 3D render using photogrammetry, unreal pipeline for large scale dataset is neat, but render pipeline is one of the worst i’ve seen in the offline render industry, it looks like it have been designed only to record game-cutescenes, but not for other type of renders.
MRQ will :
- Render everything until the start-range (including subsampling) instead starting only at the desired frame, or with a checkbox user could check to simulate non-deterministic effect (that would still run them internally without drawing them in max quality with subsampling and everything).
- Render the frame even if the lumen ligh-cache isn’t filled, render the frame even if the cesium tile loading (until reprojection error) isn’t finished, render the frame even if the lidar point cloud plugin doesn’t have finished loading it’s tiles of the octree to the number of points allowed (it’s faster to save the jpg than loading the uasset from ram to vram, mrq doesn’t care about that), mrq will render the frame at whatever cost, even if the nanite mesh is in proxy level, it doesn’t care, it render, no time for loading it to the normal amount… This can be in a way solved by prestreaming recorder, but it doens’t work with cesium/lidar-ptc.
- render black-bar if the resolution isn’t perfectly synchronized with camera sensor-size (and you have to manually calculate it)
Other things MRQ is bad at :
- you have to specify the number of foliage that could be drawn manually using command lines, same for many things that could be set in the game-override panel (it could work like the scalability menu, where you have options for each part of the engine),
- you have to pay external plugins to render equirectangular camera,
- you have to pay external plugins to render shifted lenses (camera-shift),
- the tiled render(high-resolution) doesn’t care if what it need to render is loaded (similar to point 2 above) making it fundamentally unusable to render very high resolution images,
- you can’t re-order the render-list, you have to add it by your new order manually, disabling sequence to render need to click on the dropdown button (which isn’t good design, having the checkbox directly in the list would be more usable),
- (not directly related to mrq) you can’t escape the pilot actor when you pilot camera-cuts.
I didn’t count other things like lack of clear documentation on how to use stencils clip layer to remove sky (uds) to an alpha of png/exr.