Sometimes out of focus shots need to be 'deblurred' or sharpened. This can be difficult, especially if the sequence is long and the actor's face is moving.

CopyCat, Nuke's machine learning tool, does all the hard work for you saving you time and effort.


Part 1: Deblur Out of Focus Shots Using CopyCat

In this shot, we have an actress who starts in focus, takes a step forward, and drifts out of the camera's focal plane. By the end of the shot, the actress's face is soft.

To deblur the shot, first take high-frequency details from her face on a sharper frame and put them on top of a softer frame while matching the lighting to create a deblurred frame. Then train the CopyCat algorithm on the difference between the deblurred and blurred frame.

Once the training has been completed, you will be able to deblur any part of the footage where the actress is out of focus.


Part 2: Using CopyCat to Deblur a More Complex Frame

Building on the lessons in part 1 and using the same shot, Alfie runs through deblurring more complex frames from a larger portion of the shot where the actress is out of focus.

Learn to create more in-focus training frames giving CopyCat access to more training data. This helps CopyCat produce better results.

Alfie also creates identity Frames for reference, which are neutral frames that help CopyCat during training.

This second video reinforces the skills and techniques you learned in the first video but also allows you to delve deeper into CopyCat's training.


Course Files:

Follow along by downloading the Nuke scripts for both shots in the course files tab.

The Nuke scripts are available for both Nuke and Nuke Indie.
The CopyCat training files are unavailable for Nuke as the tutorial was created using Nuke Indie.
However, you can retrain Copycat yourself in NukeX using the scripts provided.