Adaptive Subdivision Animation Option
So we all know and love the 'experimental' feature called Adaptive Subdivision were the mesh is subdivided based on what's inside the frame. This is great, but when your camera is moving, the subdivision needs to be recalculated for every single rendered frame. This leads to a lengthy render time per frame. I think there should be an option to have blender calculate where the subdivision is going to be seen throughout the entire animation, and calculate it all at once, so your first frame may take a little longer to render, but the remaining frames won't have to repetitively recalculate the tessellation.
3 Comments
I feel like this would have many way to go wrong. One of the points of adaptative subdiv being to be able to reach higher subdivisions in some places only where and when it is needed, baking everything would ruin this advantage and make monstrously heavy scenes from the get go.
You're right, but there's still a problem.. if I launch a render with persistent data active, it calculates the distance between the object and the camera at frame 1 and keeps it until the end of the animation. Even if the camera gets very close to the object, it keeps the subdivision of the first frame. You are forced to disable persistent data, but doing so at every frame it will have to reload all the textures present in the scene, thus slowing down/lengthening the rendering times... Correct? or is there a way to work around this problem?
Then maybe the proposal should be about being able to granularly chose which data should be persistent (i do believe sometimes people also have issues with external caches and whatnot). At least have a dedicated option for adaptive subdiv.