Pre production of the film started very early on this episode, around January 2015,  therefore I had time to think about a little story I quickly put on paper with drawings and concept arts first, then with the creation of a storyboard. The final scenario is highly similar to the original one, for my greatest pleasure, as I wanted to direct a real short film and not a common commercial film.

Using a less realistic character design gave me more freedom about visual creativity like in the end sequence. After years passed watching CG cinematic films from Visual Works, CG studio working with Square Enix, it was all natural I jump on the occasion to try creating a sequence that looks like those we can see in Final Fantasy games from 2000’s. With all the knowledge gathered through the years and workflows I start to fully understand and use, I wanted to have fun for the end of that Christmas adventure.

After a few drawings and some designs, Julia, the main character was born. The idea was to create a child between 8 and 12 yo, that’s why she wears a huge cloak shaped like a bell and a kawaï half bear, half panda hat.

Secondary characters all come from Julia model. I used a method I call “the Miyazaki workflow”: use the same head with different haircuts, that works pretty well. There was 4 types of characters sharing the same base mesh: girls, boy, women, man. Man based on boy’s topology and women based on girls.

Poly modelling workflow was used to create characters. All based on drawings, details were added with displacement maps created during texturing process.

Anatomy of the characters are based on Toru Ehara works. The artist found a good compromise between real and manga anatomy, therefore I used his work as inspiration to create the bodies of my characters.

Most of the film objects are based on real life objects from my house. When I missed references, I looked in my pictures database to fill the gaps.

In the 2013 episode I explored a few tracks on how to create villages for the final shot of the film. I finally used 3ds Max paint objects module, a long and destructive workflow.

This year, I welcomed RailClone! Once assets have been created (low poly Alsatian houses), all I needed to create was a single array driven by a spline to draw streets full of houses aligned to the spline segments. To test the process, I downloaded a map of historical centre of Eguisheim and drew splines on top to generate a village. The result was conclusive enough to be used in the film.

The plugin was also used to generate some wood floors and roof tiles. It became very convenient to generate medieval pavements on the roads and get random patterns in just a few clicks. A totally non destructive workflow compared to my 2013 attempt.

I definitely couldn’t manually place all the books, DVD and Blu-ray cases in the many furnitures of the film. Here again, RailClone did a great job, after watching some tutorials on Itoo Software website, I could manage complex storage equations in a few minutes.

Some elements passed by what I call a “smart modelling” step. Christmas biscuits were covered with sugar using SnowFlowPro plugin, Xmas balls were generated with procedural textures, Julia’s scarf has been positioned around her neck with a cloth modifier… Never hesitate to go deep in the technical aspect, leaving the artistic side for a few minutes, results can be surprising!

On wide shots, low poly houses were scattered with RailClone, displacement maps made the rest of the job at render time. Snow covering the models came from a material blend using a mask generated from a Groundwiz texture.

Apart from a few bitmaps, Xmas balls textures are fully procedurally generated with the help of a composite material using 3ds Max maps: Ramp, checker, tiles, etc.

That way, many materials could be created and modified in a matter of minutes.

White plastic brackets are made of a blend from procedural and normal bump to handle the deformation of thin parts of the object.

About hair process, I used the workflow established on earlier episodes, updating some processes.

Before jumping in pure hair creation, I took some time to think about the many designs of Julia hair. I 2D painted different hair styles over a Julia WIP head, based on Japanese stylised haircuts drawn by Ilya Kuvshinov.

FurGuides Painter has been my best friend on the project.

Guides has been painted over temporary hair sculpts or directly on head skulls, then for Julia and Mom Bredeles, I added Hair Farm modifiers: wisps, frizz etc. To create the main hair volume.

Using the paint of hair sculpt models method, I directly faced a problem, first vertex of each guides didn’t touch the head skull, instead of manually moving each one, I asked GJ if he had a solution and he came back to me with Hair², a maxscript fixing the issue.

On the brother model I chose a different approach. After guides creation, I added a Hair&Fur modifier to generate more guides between original guides from FurGuides Painter, this method create flat strands of hair then I add Hair Farm modifiers.

Creating hair following this method simplified and optimized the animation and simulation processes.

for 2015 episode I decided to put mocap a few feet away and go back to keyframe animation as I want to increase my animation skills. Some bad things here and there in the film, but overall, the experience was conclusive.

Each animation was shot on camera to get a good rhythm reference, I also started from the storyboard and the camera motion preview to get strong bases. Key poses have been extracted from the video to create a pose to pose animation, then I filled the gaps with keyframes and adjusted the shot length, shot global rhythm and camera path.

Hair was driven by cloth simulations. As hair are basically generated on spline cages, they were skin wrapped to low poly geometries representing the global or per strands volume.

Julia had 2 different hair model created, one for indoor shots and another for outside shots when she wears her hat. Each model were separated in many hair parts: global, strands, fringe. Working with divided parts of a hairstyle made the animation process more easy, even more when you’re dealing with long hair.

The most complex simulation was for St Nicolas hairs, at the time the model was created, in 2012, nothing was really optimized. However, the character beard simulation was so smooth and flawless, it looked like a dream to work with.

Taking most of the time to be computed, environments were the first elements to be rendered. When camera motions weren’t too complex, some backgrounds were camera mapped to avoid long render time, sometimes up to 1 week per passes. Example: 1st shot, all the Christmas market was rendered as a still image and then projected over low poly geometry, this background was then rendered in a few minutes.

This year I finally had THE workflow to render hair. During Mental Ray render passes, hair and fur were present but invisible to the camera, only projecting their shadows on other geometries. Hair render passes, 1 per light rig, were done in Scanline and then composited in the image. Apart from a few crashes in 1 or 2 scenes, the methods works until the end of the production.

Houses indoors were lighted by HDR textures.

I captured 360° HDR images that I projected on boxy rooms inside the houses models, Final gather handled the rest of the job at render time.