Deep

Using deep is as simple as connecting any deep input to the image input on the MagicDefocus2 node.

When the deep node is connected, the math option will be greyed out as this is retrieved from the deep input. The same happens to world unit and depth channel, as all information is already available from the deep input.

Warning

Deep is a heavy format and should always be pre-rendered when used.

Having more CPU threads available will make MagicDefocus2 perform faster.

Things to consider when using deep

Make sure the deep input has sufficient data. Some render engines optimize the data by removing deep data. This removes samples behind objects. While it makes the files smaller and lighter, it is not sufficient for defocussing as this data is needed.

You can detect this by using a DeepToPoints node and look at the points.

Wrong deep — First image is invalid and shows the cuts that are made in the background. The last one is good as it shows that there is data behind the objects.

Good deep — First image is invalid and shows the cuts that are made in the background. The last one is good as it shows that there is data behind the objects.

Deep slices

During processing the deep input is sliced in the z-depth. By default, when the value of deep slices is set to 1, the far field gets one slice, the midfield gets a slice and the near field gets a slice. If there are visible artifacts, it can help to increase the slice value.