Category: Nuke t1

For the final assessment we had to apply all the nodes and instruments we’ve learnt so far in the term. These included rotoscoping, colour grading, tracking, planar tracking and 2D clean-up. Having chosen this shot I took in a museum a while back, I planned out of what I wanted to clean up, what to replace and what to add.

Firstly, I denoised, then stabilized the shot. This stage was required to make sure that the tracking would be done easier and better. Then I cleaned away the picture on the left and a golden decoration on the middle wall.

Having found difficulty to grade the bits and create keys for animation of the grade node, I found that the order was important, i.e. placing the Grade node after FrameHold, before Premult.

Then I focused on replacing the art. Here the main area of problems created was the planar tracking as the pictures corners were moving randomly at few frames.

After completing the tracking and colour grading the paintings accordingly to the original shot, I found some footage of moving people that I wanted to implement inside my video. This is where I focused on the rotoscoping part of the assessment. Deciding to place the woman in the place where you would be able to see her in full size, I had to think about her shadow and how that would be formed given the lighting in the shot, arriving to the final stage by trial and error. Working on the roto of the main on the right-side was specifically tough for his hair, as there were a lot of baby hairs or space between hair at the top of the head. Overall, I like how he is facing the crowd and it makes it feel as if he is talking to the woman in front of him.

This week we were practicing the clean-up of the shot. In the shot provided, I removed the papers on the board, cleaned the side of the lockers and removed a pipe in the background.

In the next shot I removed a sand pattern, house and a car.

We had to focus on understanding the corner pin and planar tracking of nodes ‘tracker’ and ‘roto’. First, I did the exercise where we had to replace the posters in the original shot.

Looking back at the result, it feels that an additional work on getting rid of lens distortion before additions of the assets and re-applying it afterwards would have produced a better outcome. Here is the Nuke script with close-ups.

The second shot I worked on was something I shot of a camera moving into a map of the shopping mall. Trying out to stabilise the shot with the 2D tracker before applying the planar track and adding on the asset produced a result which I was not too happy with. The image is getting warped at some frames, so probably using the stabilisation techniques with application of 3D tracking could have produced a better result.

We had to practice the tracking node in Nuke for this weeks exercise. Using the footage shot handheld during the trip to St Petersburg I had last year, I practiced applying this instrument and inserted some checkerboards into the various places that were tracked.

Firstly, I stabilized the footage using 3 different points at different depth of the shot. This reduced the warping effect that occurs sometimes when applying this node. Afterwards, I chose three different windows which I would be applying and attaching the checkerboards to, finalizing down to the left, middle and right windows. The main problems that occurred were with the left window, specifically for the few seconds it would disappear from the shot: the checkerboard would not update the correct scale, as the shot is being zoomed out. I solved it by adding the keys at various frames for the CornerPin node, however, it still feels that the checkboard is changing its size incorrectly. With the window on the right side it was harder to find good and clear points, so I applied extra contrast via the grade node to the plate the tracking node was used on. At some frames the corners are jumping and I tried fixing it by adjusting the tracking keys by hand, however, some of them would move incorrectly, so it was left as it is.

For the second video I used this cool shot of the rotating ceiling followed by downward tilt. Having tried applying the tracker node for paintings on the ceiling, I decided to use the planar tracker specifically for the frames where the camera was rotating, followed by a normal tracker in the frames where there is little rotational movement. It was hard to track the 3rd point (first bottom right, then top right) of the first picture that we see. There is still movement of the checkerboard that I don’t like, but I am not sure how to solve it, as I tried re-tracking it using various points for a few times.

To enhance the tracking abilities, I applied grade and sharpening nodes to the denoised plate before using any tracker nodes.

Further practice of colour correction on various still plates. This time I chose to use the same rotoscoped picture of a hobbit, Bilbo Baggis, captured in this specifically composed shot with such a camera placement which could have been easily placed in a lot of other shots of a very similar perspective. I practiced with other 3 various plates from different moves and quite different lighting conditions.

Merging with this scene from Avatar was tough as the colours are quite blue and cold, feeling quite extreme sometimes. It was challenging to get the correct skin colour tones, with the reference used was the exposed elbow of the person lying in the middle of the street.

It was easier to match the colour of this shot from Jurassic Park.

When adding the hobbit to this frame, to add more realistic feel I added the shadow as well, but it still feels like it could be stewed more in the X direction. I also had to re-try the colour correction in order to get the skin tone correct.

For the week 5, the exercise was to roto out an image, white balance it and colour grade it correctly to fit in the scene of the second image. I practiced with both still pictures as well as moving picture.

For the stills, I chose this picture of Captain America and decided to put it into this image from “Indiana Jones and the Last Crusade”.

First, I cut out the roto shape, which I then white balanced and colour graded using the information from the second image.

Due to the position of the legs closer to the camera, I had to place the objects in the space between the legs and the background from the second shot. In order to do that, I transformed the colour graded image of Captain America to where I wanted it to be positioned in the second shot and merged together with the roto of blocking subjects in the 2nd shot using the Merge(out) node.

This is how the final shot turned out.

Looking at the final composition, addition of the shadow falling towards camera would have made the shot appear to be more realistic. Below is an image of how the Nuke Script looked like.

In the moving shot I used this poster of Uncle Sam and placed it instead of one of the pictures on the wall, following a very similar path as for the still images.

For the 4th week our coursework consisted on rotoscoping about 60-80 frames worth of an organic movement in the shot. Here are the original, green matte and with gray background rotoscoped shots.

I included the shapes around baby hairs so that I could apply the luma key node later during the course in order to isolate the dark hairs against the contrast of lighter wall. Following from the comment by a tutor, I worked on adding feathering to my roto shapes such that they would not look as sharp as before.

The toughest part of the shot were the movements of the shirt material with all the folds in various places. I found the arms to be easiest as I had a better understanding of how to divide the shapes between various places, after I had some practice with head and the body.

For my second rotoscope, I used the footage taken earlier in week 2, where I rotoscoped out the running girl at the last frames of the shot, i.e. when she is passing by the camera. I found getting the correct feathering for the coat the hardest. Another difficulty was drawing roto for the cap as it would occasionally be against the white background and it would be hard to see the outline.

Here are the roto shapes I used to cut her out.

For the second week our assignment was to shoot a plate of approximately 30 seconds long of some organic movement in the scene. Immediately I wanted to film something outside with those autumn-stylized colors. My ideas were based around two friends accidentally meeting each other outside while both occupied with various activities: one taking a walk, the other a run. Setting up the composition, thinking of the flow of movement I was happy with the finalized following 3 shots.

The first 2 shots have the same focal length of 31mm, thus producing less distortion on the sides as the characters walk into the shot. The 3rd shot was at a focal length of 24mm. The distortion at the edges is now more visible and we can definitely pick it up as the two girls walk out.

The settings were kept at shutter speed 1/30, aperture f/9.0 and ISO 2500.

One of the mistakes that I noticed only later, was that the frame rate was set at 24fps, implying that ideally, we would have wanted the shutter speed to be at 1/48~1/50, according to the double rule.

With that in mind, I decided to shoot an extra clip, set in a different settings with completely different lighting and colour tones.