Compositing Nami The Last Witch

Semester 2 Easter – Week 13 (2018/04/02 – 2018/05/14)

Technical, Artistry


Nuke organisation and workflow

The following are some ‘before/after’ examples of shots that I composited. I used a zdepth AOV to add atmospheric distance to the shots. This was mostly important as a way to make the characters stand out from the background. The zdepth with the zdefocus node was also useful as a way of softening background areas so that the focal point had the sharpest detail. I used a lot of grade nodes too to lighten, darken and change the hue/value/saturation of the characters and backgrounds. Glow nodes gave the characters softer light bounces. Alec also suggested creating a rim light using the character’s alpha channels which I’ve added to a few of the back-lit shots. Ambient occlusion passes came in the most useful for Reymond and Nami’s hands which sometimes lacked definition because of the cel shader or emissive material.

Color/mood aims:

  • Scenes 01 + 02 – warm sunset
  • Scene 03 – bright, saturated night blue and warm orange of Reymond’s light.
  • Scene 04 – Rainy, grey desaturated
  • Scene 05 – spooky desaturated to violent red
  • Scene 06 – violent red
  • Scene 07 – calmer blue


Scene 03 Púca Animation

Semester 02 Week 06 – Easter (2018/03/05 – 2018/04/04)

After Matthew had created the stepped animation for Nami in scene 03 (Reymond’s scene) I jumped onto blocking out and splining Púca’s animation. I didn’t have a lot of animation experience at this point but tried my best to apply what I had learned from Keith Lango’s workflow (blog link).


Before Feedback:

Mike gave me some feedback on scene 03 when I was part way through blocking out Púca. It was a lot to do with cinematography which has been a recurring problem highlighted in our feedback. It would definitely be worth getting more practice and research on this subject!

  • The edit has too much seeing and then saying. We need to hear and then cut to the speaker/sound.
  • Shot 02-03 Nami should go to Púca’s ear before the cut. Make sure that the camera is sufficiently different so that the transition doesn’t flash. Cut from mid of Nami and Púca to close-up of Nami and Púca.
  • Shot 03-04 action of them looking up should cue the cut.
  • sh05 Mike suggests that we go back to Púca throwing Nami off to focus on the message that Púca is thinking “crap a tree spirit”.
  • sh06 large empty space on screen is distracting.
  • sh07 when Púca looks off screen in sh06 we are cuing a subjective camera. Púca can’t be in his own pov in sh07.
  • sh08-09 Puca should start moving to cue the cut.


After feedback:

This was as far as I got with animating Púca.

Matthew did a few edits to Púca after this when he was polishing Nami such as when Nami pulls Púca’s horns at the start. My animation on Púca still feels quite mechanical in places. I think that this is due to my eases and timing which could be less abrupt. Púca’s walk was the most difficult part to animate but Alec gave helpful advice with regards to moving forward with the hip control instead of the global transform.


I mostly used the dog walk cycle in “The Animator’s Survival Kit” as a guide but with Púca’s forarms in more of a gorilla pose. Matthew also advised me to look at lions and apply how they curl in and flip up their paws to Púca’s hands as he walks.

Animbot was useful too, particularly for creating selection buttons and using the tween machine slider for holds and offsetting inbetweens of different body parts.



Reymond Blend Shapes

Semester 02 Week 03 (2018/02/12)

Technical, Research and Artistry

I sculpted the blendshapes for our character Reymond. I hadn’t created blendshapes before but Andrew had some great research at hand to help me out. He kindly shared a tutorial with me that contained a detailed workflow of all the different sections of the face that would need to be sculpted. This was very helpful as I had thought to approach this by copying the faces in the animatic which would have been slower to plan out.

This tool, abSymMesh, was recommened in the tutorial as a way of checking if the mesh is symmetrical, which it wasn’t! Making Reymond’s face symmetrical was more time consuming than predicted as mirroring one side of his face over was tricky to reconnect.

Along with the list of shapes in Andrew’s tutorial, I mostly followed along with this tutorial also using the shape editor in Maya 2017.

Maya 2017 Shape Editor

These are the resulting shapes after sculpting and duplicating/mirroring to the right.

Andrew is going to complete Reymond by attaching the controls to the blendshapes and adding more body controls. The mouth will be driven by joints more than by my blendshapes.

Rendering and Compositing Look Development

Artistry, Technical and Research

Semester 02 Week 04 – 05 (2018/02/19 – 2018/03/04)


These are some renders and composites from the start of semester 2 to get an idea for what the characters might look like in scene 03 – the road to Reymond. It still needs a lot of work. Reymond wasn’t casting enough light on his surroundings with the emmisive set -up so I’ve tried making him a mesh light in the images below and added a glow to his emmisive pass in Nuke.  This causes bad light bleed around his mouth. I’ll try light linking for his cast light.



Research and Technical

Andrew found this cel shader script for Arnold 5 which works much better than relying on hard directional lights! It’s free for download here:

The light and shadow color are controlled by a ramp. Kerry and I are setting up the cel shader in each individual asset file. A little extra work is therefore needed to reconnect every referenced asset in the layout files to the same ramp so that light and shadow HSV can be changed from one control point. This shading network also doesn’t seem to have an imput for height maps unless we reroute to an aiStandardSurface.

Cell Shader Tool – Arnold 5 – Maya from Steffano Richi:


These are some of the updated assets that I worked on with the cel shader. We’re trying to keep our asset textures broad, abstract and brushy so that they read like a painting from afar, especially when combined with the cel shader which simplifies the shading. I’ve also made a warm and cold version of the tree textures to help with our scene palettes.



This is our progress so far with rendering and compositing our characters in an environment. This is in layout 02 which Kerry was working on. The cel shader doesn’t work with the default AOVs from Nuke so this is only composited with the zdepth pass. It needs more work. Mike has also pointed out that we need to stage the characters with the lighting more.  This seems tricky with our current light-set up and workflow so it needs more research and experimenting.



I’ve started texturing the inside of the house while Kerry is working on the house exterior this week too. I was worried that the cel shader only seemed to work with directional lights but I’ve found that it works with the aiAreaLight also. Light will be coming from the doorway and boarded windows at the front and from a caved-in roof at the back.



Alec and the others have suggested using geometry with alpha gradients for light beams. I’ll also be developing the red strangling version of the house interior light/textures.

MASH Research

Week 18 (2018/01/22)

Technical and Research

These are some of the videos which I found useful when researching how to use MASH for set dressing.

MASH – Placer Node

This video shows how to scatter geometry using the brush-like placer node. I found that MASH only works in cm. There is an option to set “strict no collisions” but I found that my geometry was still colliding no matter how I played around with the settings. It therefore took a lot more care in placing geometry than what is shown in the video.


Set dressing using MASH Placer Node & Arnold Procedural /Standins

This video introduces the workflow of scattering low-poly geometry with the mash placer node which are then swapped out at render time with the detailed versions. The arnold help page on An Introduction to Stand-ins shows how to export geometry as stand-ins and be re-created as bounding boxes in the viewport. So far we haven’t had the need to use this method.


Maya 2017 – Intro to Mash

This video introduces how animation from an origin mesh can then be duplicated and controlled with a time node. This node is only available with non-instanced MASH and therefore requires a low to high number of meshes workflow.

MASH – World Node Clusters

I looked into this as an alternative to using the placer node. An option in the World Node allows you to generate clusters of geometry around the first geometry ID in your network.

Reymond Textures, Leaves and Trees

Week 17 -18 (2018/01/15 – 2018/01/26)

Technical and Artistry

I tried to match Reymonds textures in Substance Painter to my previous concept art of him. We’re trying to keep with Reymond’s scene being a mix of warm oranges from Reymond and cold blue moonlight.


These textures might need more work if they make the facial expressions too flat. Alec suggested using an ambient occulsion pass to help define features.

For Reymond’s leaves I modelled one branch of tertiary/smaller branches and leaves and then distributed this using the MASH placer node. MASH is somewhat fidgety in that it doesn’t always distribute geometry near the cursor or if I create a selection set of geometry it sometimes only distributes across the first few members of the set. I’ve also had trouble with getting the MASH network branches to move with Reymond’s geometry. There might be a setting to fix this in Maya 2018 but for now we’re doing all of our setups in 2017.


I tried adding a small amount of animation to the MASH repromesh by animating each of the original leaves. Animations can be easily looped, offset or scaled by using a MASH time node. This node can only be used on MASH copies (repro) rather than instanced meshes though. My result is a little too subtle but it’s something I’ll come back to if we have time to add environmental animations.


In semester two, week one I mostly concentrated on trees before Kerry and I started a full asset list for proxy models and set layout. I finished off the tree models that I had started before Christmas, UV unwrapped them and then textured them in Substance Painter.



I tried a different approach to leaves with groups of leaves on an image plane instead of every leaf being on one plane.



I tried using a blue texture plugged into emission to control the shadow color.


Reymond Progress

06/Jan/2018 – 15/Jan/2018 – Technical and Artistry

I’ve been tasked with modeling Reymond. Below are the references that I was working from, from Andrew’s drawings and some of my own sketches from before Christmas. I had painted Reymond as a towering tree but Mike pointed out that Reymond’s framing would be completely changed with this, changing the comedic audience eye level in the animatic. I’ve therefore resized Reymond’s model to be more at an eye level with Nami and Púca.


Andrew’s drawings of Reymond from the animatic.



I haven’t modeled any facial topology since first year so I started by looking back on my research from back then.

My previous research on head modeling: link

The topology reference from hippydrome came in the most useful:

I choose to start with topology in Maya instead of organic sculpting. I drew out where I thought the edge-flow would be based on the reference from hippydrome. I tried duplicating this by moving edgeloops on a box at first but then I just found it easier to trace my drawing with the quad draw tool. This made a flat plane of facial topology which I then pushed, pulled and pinched with the Maya sculpting tools until it looked like my drawing. I tried to make him chiseled like in the sketch by pinching the topology closer together in places but I hope this doesn’t cause any problems with the facial blend shapes and rig.


Work in progress of Maya sculpting:

Reymond progress turnaround closeup:

Reymond closeup with mesh:

Reymond progress turnaround:

With mesh:

I still need to add more hanging moss on Reymond’s branches and I’m going to experiment between Substance Painter or sculpted displacement maps to create his clumped moss.

Reymond renders:


I tried to create Reymond’s UVs in as automated a way as possible, as in my previous blog post rendering a tree. The unfold and layout tools have been working a lot better in Maya 2017 than they have before. I wonder if this is because I spent time using the mesh cleanup tool?


Look Development

Week 15 – Friday 2018/Jan/05 – Artistry and Technical

We’re trying to figure out how our cartoon world will be textured, shaded and composited. I tried making some simple textures in Substance Painter and brought them into Maya with the aim of combining them with a cartoony shading and lighting method.



I tried applying toon shading with the facing ratio shader to my tree model (Arnold tutorial) but did not like the result as it doesn’t pay attention to my light setup.



I’ve been trying to figure out how to render lights separately in Arnold for a long time and it turns out that it’s extremely simple. I found instructions for using AOV light groups on the bottom of this page:

By rendering lights separately we could maybe play with the contrast in post for getting the look that we want? Somewhere between cel shaded and having soft glows in places. This is what I’ve got from Nuke so far by rendering direct and indirect AOVs for three lights: dome, key directional and rim:



This tutorial was useful for UV unwrapping the tree. It makes the process as automated as possible, using planar mapping, auto-seams, unfold and layout functions.


Andrew has shared this rendering tutorial with us which will hopefully help our look development.

Week 10 Maya Lighting, Arnold AOVs and Nuke Compositing

For week 10, the team divided into two segments. Matthew, Sorcha and Andrew focused on the animatic while Kerry and I focused on trying to recreate the house scene exterior. I blocked out the scene trying to match the scale and perspective/focal length and composition of the concept and then shared the project with Kerry. Kerry focused on modeling the house and bridge while I focused on modeling the trees and testing the lighting, rendering and compositing.


I was having trouble getting the Z depth pass from Arnold to work in Nuke so this first render test was created by rendering the geometry in separate render layers in depth and then manually adding the atmospheric depth on each layer in After Effects. This resulted in a lot of control over the layers but created more work in setting up different render layers. I think that ideally we’d want a lot of control but not as many separate layers.



AOVs and Nuke

I was loosing hope in using the Z depth pass. I had got it working before by grading the Z pass in Nuke so that it was visible in the viewport but hadn’t tried it in a long time so was missing a step. Alec as usual was lots of help!:) He went through an Arnold to Nuke workflow with Kerry and I and showed that the Z pass could be read by a zDefocus node in Nuke for simulating depth of field. Alec also showed us the benefit of having separate AOVs that could be edited in Nuke.

Andrew shared this video from Arvid Schneider’s YouTube that goes through the same process of rendering multiple Arnold AOVs into one merged EXR file and then extracting and recombining these in Nuke with shuffle nodes. This video also shows how to use the Z pass to create atmospheric perspective (finally!). The Z pass needs to first be extracted from the merged EXR and then the depth channel 0 to 1 space needs to be normalised by grading the white and black points with values from the clipping plane (if I understand that?). Schneider also shows a useful way of creating a custom AOV for material IDs using the aiWriteColor utility.

Arvid Schneider MtoA 120 | AOVs for Comp | using Arnold with Maya 2017



It took me a while to understand the different steps in this workflow. This page has useful information on the different sets of passes/AOVs that can be used to compose the beauty pass.

Nuke swapping channels

I got the composite working for a simple scene with spheres. Alec showed us a possible method of creating light bloom by rendering a rim light in Maya with the indirect light turned off and then blurring this in Nuke to get a soft light halo effect. I added a separate render layer for the rim light’s direct light as I can’t yet figure out how to render different lights into their own AOVs with attribute overrides. Creating light bloom needs to be further looked into. Alec had advised to work towards making rendering and compositing templates that could be used as a base across scenes.


Alec suggested pushing the atmospheric fog and cinematic composition even more, as in the last image. I also made the mistake of making the depth of field too strong so that it looked like miniature photography. Sorcha pointed out that dof should be used for objects close to the camera. I had a feeling that this was the case but I haven’t experimented much before with where the blurring starts and stops with objects of different scales.

We haven’t started experimenting with any textures or shading yet. The light in the window is just a temporary Photoshop layer. Alec suggested creating light scattering with either environment fog or faked god rays wrapped around geometry with an alpha channel. The grass at the moment is also just simple Maya Paint Effects so this might change with the texturing process. The rain and fog are Photoshop overlays so will have to be looked into further also.


Technical – Plan

Even though we’re developing our Cassandra Cottage script to have a largely different twist than what we first planned for, a lot of the technical challenges and risks still remain the same.

Animation – We want to push how cartoony and energetic our acting is. I admit that I don’t even know where to start with doing this besides timing, squash and stretch and smear frames so there’s a lot of research and critical thinking to do! I feel like I need to recap on the basics of animation, from ball bounces to creating different characterful walks, as my focus for this past year has been on lighting, rendering, texturing, grooming, dynamics and concept art. Even if I don’t contribute a lot on the animating side of the project I’d still like to know enough to give constructive feedback, particularly on performance.

Rigging – So far we have our three characters who need rigged (two bipeds and one quadruped), Púca’s bird form and then the monster in our episode (which may be an angler fish like monster house or a bipedal sphynx).

Style – We seem to be gravitating towards more toony lighting and shading with sharp edges. This is something which we’ll have to experiment with across modeling, lighting, shading and texturing. We also want to capture the look of light bloom (if I’m using that term correctly, bright feathered light bouncing from edges) and we’re not sure how to do that yet. Regardless of what style we use, we will still have to consider how best to set up and composite the layers for atmospheric perspective, depth of field/z depth and color gradients in depth.

Sound – our short has a lot of dialogue. Not only do we need to make it look good on paper but we need to find and direct voice actors to get the best performance for our short. We’ve also been discussing how we can use music in our short and like the idea of adding more rock and metal elements.

Fur – I’ve used XGen before for grooming lots of fine fluffy fur. I didn’t get as far as making it work with zync or RebusFarm because of errors so this is something which we’ll have to troubleshoot if we go down this root. We’ll also need to consider how the fur design fits in with the overall style of the world. It could be as simple as abstract textures or as complex as shaped geometry. Looking into making the fur dynamic is also going to be a challenge, even if it’s just subtle movement to make the character more alive.

Shapeshift – Our Púca character is a shapeshifter and we feel that it would be cheating an important part of the show from the audience if we didn’t show it at least once in the short. We’re planning on looking into compositing tricks but perhaps there is an inbetween form that could be done with blendshapes. Maui from Moana shapeshifts in a frame which could also be a simple option. Aku from Samurai Jack is almost like dynamic black fluid between forms. It’s something that we’ll need to experiment and do concepts for. Our new idea may also involve a monster house with transforming parts.

Magic and explosion fx – We’re considering how we could stylistically fit in the magic visuals in our short. Will it be 2D shaded dynamics?