George Quelch - VFX

Something I have learnt whilst studying and working is the importance lens and film stock choice affects the overall look and feel of your image, and I think its really important to carry this over into fully digital work, or matching it to backplates. A big part of my learning more about this was my LensFX gizmo . I improved it a lot during the last few months of uni to get it to a place I was really happy with. Adding features to add lens dust and focus breathing. Below is a test I did in Arnold, playing with some displacement, trying to emulate the features/effects of a real lens with focus breathing. I added a focus pull, lens breathing, a lens flare, lens dust, filter flares and subtle glows. To make it feel more cinematic, I worked at 1920*800 widescreen resolution and created anamorphic style bokeh. I think it makes the image feel a lot more dynamic that if it was a simple focus pull, with no breathing or refocusing.  Best to watch this video in fullscreen at 1080p ...

Attribute controlled Shading In order to further my rendering and shading technical knowledge I have been exploring Primitive Variables (Renderman) and UserData (Arnold) attributes in order to set up more efficient Maya scene files. The goal here is to cut down on the number of shaders used in a scene by having only one across multiple objects, but still having variation via custom attributes on each objects shape node. A very basic example: you have 3 spheres, a red, a green and a blue. In a simple setup each might have its own shader, with it's specific colours.  However using PrimVars, you can have one shader for the spheres, and each sphere would contain an attribute with its own colour value. At render the shader gets the colour value from each object, and produces exactly the same result, but all from one shader. Learn more about UserData here This is gonna solve optimization problems on Dino Diner, where we may have 10 tables in a scene, eac...

Towards the start of 3rd year Cherry Cherry VFX provided us with the oppertunity to work on an outside brief. I worked as part of a small group of students over the course of Semester 1. My roles were lighting, rendering and compositing. We had to come up with an idea for a popup cinema sting, a student put an idea through and after some deliberation with the client w e were given a brief to create a seagull floating on a life buoy in the open ocean, and a clapperboard jumping over, in full un-stylized CG. Starting the project we knew we were going to have to blend together simulated water in the foreground and an un-simulated background water, which we could stretch out into this distance. Also creating a good bird rig was going to be a challenge, as well as achieving a photoreal or close to final result. We ran into a few issues with this project, in several different areas that had to be overcome, for me these were mainly to do with the rendering of the wat...

Now  that dissertation is finally handed in I have been able to put more time into working on compositing for Dino Diner. After the many tests I did for the original previs model here  I wanted to take what I thought was the best technique and apply it to the new previs model.  From Arnold From Nuke I opted for rendering no volumetric lighting in Arnold and added them using Nuke and the TX_Rays node as with my last previs. I also made use of Normal and Point world passes to add additional lighting to the table and counter behind. I also added floating particles which interact with the light shaft, and a range of camera lens effects, including chromatic aberration, depth of field, blurring and more. This time I kept the additional lighting to a minimum, as in my last test I found it caused issues with matte lines and aliasing around edges with additional lighting. I think this issue could be somewhat resolved by making use of several matte IDs for...

A side to looking into Volumetrics in Nuke I have been exploring the Depth to Points node. It's a very clever workflow that only needs a depth and normal pass, which allows you to project your render out into 3D space within Nuke. This process would save having to export geometry and cameras out of Maya, it would all be self contained within the render and the passes, which I thought was really interesting and convenient. To give this technique a test I decided to composite in a sphere into the background of the shot. This is what I was able to produce. Which considering was all done in Nuke without and interaction after render with Maya is pretty cool. I was able to position the sphere within the projection, then by looking in 3D space was able to match where the lights were in order to quickly match the lighting on it. I also tested it with volumetric lighting, although this caused me a lot of trouble with matte lines around the table, I think ...

This week I have been looking into different render solutions for volumetric lighting in 3D. I have been working with Arnold to create some basic lighting with a previs model for the cave created by another student. It was only a quick render, whilst I felt this worked OK the volumetrics were quite noisy and added a lot to render time. I am confident with adding them into 2D space with Nuke but had not added them to 3D successfully yet, so they work with a moving camera, so this is what I wanted to look into doing with Nuke. I found a good toolset with the TX_3DRays gizmo from Nukepedia. This allows quick creation of a light shaft which you can positing within 3D space, by reading in the geometry I could also use it as a holdout so the volumetrics sit properly within the scene. Using two slightly different workflows I created the two images below, which have slightly different looks. I prefer this first image as the result is more subtle, and i...

I and several other students are working on a side project for Cherry Cherry VFX. We are creating a short cinema sting, featuring a photo-real seagull, floating on a lifebuoy in the open ocean, and an clapperboard jumping out above it. There was a lot to figure out with this project, mainly with the ocean as we need it to be simulated to create the splash, we need it to base animate the motion of the life buoy, and also be able to render it as a wide open ocean. My role is mainly to do lighting, rendering and compositing, and have been helping with pipeline, especially on the water simulation. Harry has been working with BOSS and Bifrost to create a workflow in which we can drive the bifrost simulation near the camera using a BOSS ocean, which means we can make the splash. But further away from camera will be a tiled version of the base BOSS ocean, without the simulation. We were struggling with long render times on the ocean as it is simulated and quite dense. We ar...

For look development on The Big Top I am continuing to look into cloth shaders. Again testing in Arnold. The idea behind this was similar to what I did previously with Renderman. Using maya's cloth texture which I adjusted to my liking, and used it for a bump on the model, which was simulated with nCloth. I also added the small XGen hairs to enhance the fuzzy effect, I think this worked well. One thing I want to look into doing with this is small bumps, and caught hairs you get across different fabrics. One thing I did differently was use several noise textures to get variation within the colour of the cloth, so it wasn't one perfect shade throughout. I want to get a better look development maya scene set up so I can work on things in better lighting conditions, also so it is universal between engines.

A big part of the pipeline I designed for The Big Top was using HDRs of the sets for lighting the CG characters. This makes the process of lighting the CG much smoother, and the end result will look better. We shot 9 exposures from 2 angles using a Canon 5D mkiii, and a 300mm lens.  One of the challenges with shooting HDRs for stop motion sets is the size, so we had to find a camera with a short enough focus distance to capture both angles in the small space, but long enough focal length to fill as much of the frame as possible. Once we had the images I used Photoshop's HDR Pro function to merge them into one single image which I exported out as an EXR. I then took this into Nuke, to do the unwrapping process, I did this  with the spherical transform node, which worked well. Obviously there is massive distortion to the left of the image, and to the right you can see where we were taking the photos, this is where the second image is needed. ...

For look development on The Big Top I have been working on shader tests for plastic materials, in both Renderman and Arnold, these first 3 are rendered with Arnold. We are taking reference from Evil Corp  who make great plastic style CG models that look very realistic. The key things I picked up on was the subsurface scattering, and the amount of specular on the skin, so before we get some models and textures I am experimenting with these properties on this model. I played with getting the scattering correct, which you can see working on the ear. I then played with varying amounts of specular and reflection. I don't think any of them are quite right yet, it will be easier to play with shaders once we have some textures painted. I do like Arnolds ability to add lights for specific attributes, for example subsurface scattering, although its not physically plausible sometimes you want that control to push an effect in specific areas, although this could probably be...

For R&D on The Big Top I have been looking into different ways to create crying in 3D. The first way I tried was hand animating a single tear, or several if needed, and creating the streak effect left by a tear within Nuke. It's a simple way of working but the technique worked. It also allows us to control exactly where the tear flows down as it is hand animated, rather than a simulated effect. I rendered the Face, the face with a high specular override, and the tear as 3 separate render layers from Arnold, using Maya's new powerful render layer setup. Once I had these I took them into Nuke. I fiddled with Time Offset, Time Blur and Time echo before deciding the latter did what I wanted. This let me make the streak which the tear leaves, I graded and blured this before merging the original tear back on top. I then used the resulting alpha to merge over the specular pass of the face, giving what would be the water left by the tear some reflection, I think this real...

The past few weeks I have been looking into XGen for maya properly. As my past experiences with it have either been brief or it's been a nightmare to work with. This is partly to help with look Dev, and being able to take shaders a little bit further with some hair, and the 3rd year project Dino Diner, which may need fur simulations. I've made good progress with this tutorial . I now feel like I have a good understanding of the grooming tools, and the extra detail that can be added via modifiers, how the groom controls the primitives, and how these get be rendered with custom shaders.  It was complicated at first getting my head around the texture maps to Ptex conversions but was more simple once I understood what was actually going on and what XGen was actually doing. I'm pretty happy with the result, I would definitely spend more time on the groom if I were doing this for a project, to get all the hair flowing properly. This was rendered ...

Having completed my game of life script I wanted to get a nice render of it, which I've finally gotten around too. I wanted to make a grid of marbles, each different colours, however having around 2000 in total in the scene that was not gonna happen by assigning them manually. After a lot of trial and error I found a way to get what I wanted. First I created a procedural texture of squares, each different colours from the 6 base colours I chose. I then positioned this under the spheres. I gave every sphere a completely see through glass material, and disabled primary viability of the plane. This meant the refracted colour of the plane could be seen in the marbles but not the marbles themselves. I think this worked pretty well, definitely more efficient than  assigning  all the objects a material.

After a summer that flew by taking part in the BFX Festival Competition, I'm trying to manically work through my list of saved links and techniques in preparation for 3rd year at Falmouth. I'm hopefully going to be able to talk more about BFX soon, including the film we made and some of the breakdown, and work I put into it, but we are currently waiting for the films to be released online to be able to talk about it. For now I am focusing on getting to grips with Arnold, and Maya 2017 in general. There are lots of little improvements all around, especially with the animation and general interface changes. As part of a continuation of the lighting I was doing before I took the  St anford dragon  into Arnold and lit it with a HDR image. One of the big problems I had with Mental Ray in 2016 was rendering  caustics , this however worked quite nicely in Arnold. I was pretty impressed by the quick results. The glass looks cle...

I have lately been interested in trying to learn more about different effects to add to CG composites in Nuke to give them more real world properties . I rendered all these images in Arnold, using a HDRi and key light, then took that into Nuke, and did a rebuild of the passes to get the beauty. The background was added on using the HDRi mapped onto a sphere in a 3D scene. One thing I was interested in trying was different filters on the ZDefocus node, this can give a little more interesting bokeh, that of you would find with using different physical camera lenses. Here is a short article about it. This is the basic ZDefocus node blur, with no changes to the filter: Different camera lenses can produce differently shaped bokeh, due to the number of  aperture  blades within the lens, for example, 3 blades produces triangle blurred highlights: or 4 blades produces squares: The ZDefocus node has 3 filters; disc, which is most ...