bigcrownobject-blog

[bigcrownobject-blog]

VFX 1 Blog

In this blog I am going to give a breakdown of how I achieved my final pieces of work and run through the techniques I used.

Car colour correction.

First of all, for this footage two reformat nodes are needed and we also need to change our project settings, the output format is going to be 1920x1080 on the project settings and we need to reformat both of our image which are in the scene to 1920x1080 as well. We will then begin by merging the car layer over the icy road.

Next we are going to read in the windshield image which is an alpha image, we are going to use this image to punch a hole where we want in the foreground image. To do this another merge node is added between the car image and the first merge node. 

We now go to our new merge properties and change the ‘operation’ tab to stencil, this is where the hole is punched in our foreground. A transform can now be placed after the car image which allows us to move the car to a position which we choose.

In order to begin colour correcting I have dropped a grade node under my original car image however since we cannot colour a premultiplied image we need to add an unpremult node before our grade and then add a premult node in after the grade in order to premultiply the image again. We now have full control over the grade.

We are now going to use the curve tool from the image toolbox in order to see the lightest and darkest pixels on the image. The curve tool is dropped in and is linked to our car image and the curve type is set to max luma pixel, a constant node is then created which is like a solid and this is merged with our car image as well.

Once we had clicked go on our curve tool it analysed the lightest and darkest pixels as I previously mentioned, we then opened up our grade properties as well as the curve tool properties and we are going to copy our maximum luminance values into our whitepoint and our minimum values into the blackpoint of our grade.

We then go and add a panel called pixel analyser where our properties usually show up for nodes, the pixel analyser allows you to drag around the screen and gain information about the pixels and the brightest and darkest points.

Once we have all of these values we can again copy them across onto our grade node in order to match the colours correctly.

 We also need to match the mid tones of the car and the background which needs to be done by eye. First off we lower our gamma slightly and then adjust the multiply option under our grade node until we get the desired look. There is an improvement however now the car looks too warm for the surrounding area.

Finally we match each of the colour channels manually until we get the desired look, we hit R, G and B on the keyboard separately to control the relevant colour channels and we can then edit our multiply, offset and gamma.

Laptop Tutorial

 The footage for the laptop is read in at a 1280x1080 resolution, this needs to be addressed by firstly changing the project settings format to 1920x1080 and also adding in a reformat node and changing the resize type to distort with the format again being 1920x1080. This will change the footage to the desired resolution.

The screen then needs to be rotoscoped in order for a keylight to be placed in which is how we are going to remove the green off the screen and place and track our own image onto the screen.

Once we have rotoscoped the screen we add in a keylight node, the colour picker tool is then used to pick the green colour on the screen. Using the alpha channel, we can see our key properly and use the properties of the keylight to adjust the key using the clip black and screen gain.

We now add a track point to each corner of the screen using a tracker node, this is done specifically so we can corner pin an image accurately to the screen later on.

Next we need to read in the image we are going to use over the screen. In this case we are using a screenshot of a desktop, this needs to be reformatted to fit the 1920x1080 resolution.

Now in order to give the impression of the image being on the screen we need to use a corner pin node, we can then link this to the track points we created earlier as seen below. This tracks the corner of the image to each of the points allowing the camera to follow the screen along.

Once this was done we added in a shuffle copy from our keylight going back into the source footage, this brought back some of the colour information we lost when doing our key. This was then merged with the corner pin and a premult node was added in order to remove the green overlay on the screen of the laptop and we finally had the image merged over the laptop screen.

The hand was originally going underneath the image however this was resolved by going into the keylight node and changing the OutM component to ‘none’.

 The footage was completed with some colour correction, mainly to fix the green glow on the hand a hue correction was used to supress the greens a bit as well an overall colour correct and grade to try make the piece look more natural.

Tears of Steel/01_2a

First of all we need to change our project settings output to 1920x1080 resolution, a reformat node is also added in order to change the footage resolution to 1920x1080.

Next we needed to rotoscope the fence since the robot is later going to be placed behind it, I added in a roto node and animated it slightly with the movement of the camera.

Next a merge node was added and the mask pipe was taken from the fence roto and placed into the merge node. Another merge node was added in and linked back to our original footage with the B pipe and the A pipe was connected to the other merge node.

Next the robot image was read into the node tree and attached to the merge using the A pipe to layer this in behind the fence.

A reformat node was used in order to change the resolution of the robot to match the rest of the scene, a transform node was also used in order to scale the robot in the desired position.

A premult node was then added before I merged the robot back with my roto to get rid of the white background behind it.

The tracker node was added in and connected back to the source footage in order to see where the tracker needed placing. The tracker was tracked forward from the first frame to the last, a transform node was then linked back to the robot image and then this transform node was linked to the trackers movement. This allowed the robot to relate to the transform node which related to the movement of the tracker therefore the robot moves naturally with the camera movement.

When the robot and the footage were merged it made the footage overly saturated, to correct this a grade node was added after the footage and its reformat in order to bring the blackpoint and white point up or down until we got a closer look to the original footage.

An overall colour correct and grade was added until the desired look was achieved as seen above.

Tears of Steel/08_4c

First of all, we need to start by changing our project settings so that our output settings are 1920x1090 pixels. A reformat node was also added to correct the footage to 1920x1080 pixels as well. Once this was done we began with the rotoscopes of the different body parts, five were done in total to try and get the most accurate results.

We then piped the rotos into keylight nodes in order to use the colour picker tool to being keying the green screen out of the footage. The roto selection was inverted and the keylight OutM component was set to alpha, this displays only what is inside of the selection allowing for a more accurate key. The properties of the keylights were then adjusted such as the screen gain, clip black and clip white until an accurate key was accomplished.

Some time was spent re-adjusting some of the rotos such as the one above since there were some elements such as this red sticker which needed to be removed from the final composition. Once this was done multiple merge nodes were added in order to tie all of the rotos and keylights together which gave us a full body rotoscope of our actor with the green screen in the back completely removed from the scene.

Next a background was added in and as well as a transform node to scale the actor up. The background was read in and reformatted to fit 1920 x 1080 resolution, I merged this with all of my rotos and keylights by adding a merge node, I dragged the A pipe into my merged keylights since this is going to be the foreground and the B pipe into my reformat node for the background image.

This created the effect of the guy being layered over the background image of the boat however I was having an issue where parts of the actor were transparent.

I then wanted to tone down some of the colours such as the overexposure and overly saturated parts on the actors head. I added two hue corrects to take the red out of the actors face as well as removing a blue tint under his beard. I then added a colour correct and pulled down the saturation and contrast to remove the highlight on the actor’s hair.

I then added in a grade node and a hue correct to darken the character overall since I was going for an overcast and gloomy look with the background so I wanted the actor to match the scene correctly.

I added in an overall grade to again darken my footage as a whole, I also added in another hue correct node and brought down the blue and the red colours as much as I could, I then pulled in more green colours in order to get this gloomy, overcast look. The shot ended up looking quite cinematic and fits the pirate theme I was going for well.

Window Close Push

To begin with when we read in the footage it comes at a resolution of 1280x1080, this need changing to 1920x1080 since this is the resolution which we would like to export at. We address the resolution issue by firstly going to our project settings and changing our full size format to 1920x1080 and also adding in a reformat node and changing the output format to 1920x1080 as well, this gets the footage to the desired resolution.

The parts we want to keep in the footage now need to be rotoscoped in order to remove the green screen surrounding the pieces we want to keep. Here the tree, candle stick and window frame have all been rotoscoped in order to remove the surrounding green colour. This technique is used a lot in modern day films albeit on a larger and more complex scale, a great modern day example of this is with ‘The Avengers’ franchise as those films are very reliant on greens screen techniques.  

In order to remove the green colour from the background of the footage the keylight is selected to display the properties of that node. The colour picker tool is then selected and holding CTRL + left click we can select a colour which we want to remove from the footage, we set our keylights OutM component to inverted alpha so everything on the outside of our selection is made transparent, in this case the surrounding green area. We then adjust our keylight settings in order to refine the key, this is done by adjusting the screen gain, clip black and clip white settings until we get the desired key. This is repeated for all three rotos therefore three individual keylights were used.

Next we would like to merge the three keylights as well as start to bring back some of the reflection on the window behind the tree. In order to do this, we needed to add two merge nodes; one to connect two of our keylights together and one to connect the remaining keylight to the other merge essentially merging it with the other two keylights. Once this was done we needed to add in yet another keylight which is to be our outer keylight, we then took the bg (background) pipe from all three of our rotos and linked it into our new outer keylight.

We then can go and use the colour pick tool again on our new keylight properties and in the same way we did for the keylights earlier; remove the background. We need to bring the screen gain up again as well as the clip black but this time the clip white was brought down and the screen dilate was brought up as well bringing back some of the reflections on the glass from the original footage we read in.

A tracker node was added after the viewer was piped back into the original footage, three track points were then placed on the track points you see below in order for the background to move with the camera giving a smooth natural look.