: Crimes of Grindenwald

Nov 16, 2018

Fantastic Beasts II, The Crimes of Grindenwald was an interesting show. On the one hand I was living the dream of working on another Harry Potter film (I’m a fan boy), but on the other hand, the work was complicated and difficult to manage. I was playing compositing supervisor at Framestore Montreal.

The Montreal studio had undergone a large expansion the previous year (2017). After years of preparing, we doubled in size. Dispite the exhaustion of 2017 with it’s larger shows and crew sizes, 2018 came around just the same, catching us off guard and breathless. The Crimes of Grindenwald VFX work was planed over 1.5 years in advance, so ready or not, we had to rally the troops once again, and keep producing.

For me personally, one of the interesting things about the project was our efforts to start instituting the next generation of leadership in the studio’s compositing team. While we didn’t have an official job title or contract yet, the success of Bladerunner, and the organization of the artists, allowed me to test those ideas once again. We trained 4x new compositing leads on the film.

The idea of training new leads came about because the show was slightly backended. Meaning that the shots were not able to be ‘unlocked’ and worked on, untill the edit was completed. Reel by reel, the deadlines slipped, leaving the majority of the work for the compositing team to be completed in a very short amount of time just before final delivery. This meant that our normal schedual of daily review time would not be sufficent to complete the show. There would physically not be enough hours in the day for either myself, the CG supervisor, or the VFX supervisor to review the work being created by the large compositing team. The comp leads were going to play an integral part of the production. Running sequence desklies and dailies. Taking on more pipeline and show setup responsibulites…etc.

Since it was unusual, it was a hard sell to the studio. I had to justify taking these talented artists ‘off the floor’ and place them into demi-management positions. There was a cost to that, and the production design was (appropriately) challanged. But in the end, the internal comp lead ‘training program’, meetings, and production design paid off. The artists had the speedy feedback mechanism that I wanted. The Production team had the turn around time for shotwork they expected. Most importantly, I felt like I was being honest with myself about how much I could personally accomplish in a given day/week/month. Not overburdening my self, meant I could be more available to look forward trying to stay a delivery ahead of the working team, and respond quickly to the needs of the leads/artists when they needed it most.

Picket - Ministry of Magic

Sequence Lead: Jonathan Turner

One of the more straight forward sequences on the show. Even so, there was still a lot of prep work to be done (rotoscope and paint) to prepare the plates to recieve our little Picket.

The house elf shots were multi vendor. Meaning we needed to complete the work early, and send to another vendor (can’t recal if it was DNEG, MPC or Method), so they would have time to place in the eagle and set topup.

The button shots are full CG! Luc Girard was the groom lead. I think that cg carpet looked amazing.

Paris - Circus Archanus - Night Exterior

Sequence Lead: Filip Susteck

This sequence was more indicative of the work for Fantastic Beasts II. In each shot there was a laundry list of possible elements or specific work to be done. At the time we didn’t have in place a good system for managing this cacophony of shot specific work. The large amount of creative variation between shots made quality control difficult.

Example: “Was this the shot with the Kappa & Test Darkening? Does this need off screen Firedrake sparks?” Etc.

Here is a list of the various elements which might be in any given shot:

Kappa (cg creature)

Oni (cg creature)
Firedrakes (cg creature)
Graphorn (cg creature)
Zouwou (cg creature)
House Elf (cg creaure)
Bubble Kids (plate element kids in floating bubbles)
Magic Fire Guys
Magic Animated Posters/Signs
Magic Fireworks
Magic Bronse Statue
Magic Portals (at end of streets)
Magic Carnie Stalls/Vendors/Carts
Magic Animated Whistles
Place Cache Set extension
CG Circus Tent(s)
CG Balloons
CG Cotten Candy
Set Greeking (because it was used in other scenes)
Set Continuity plate clean up moving targets as cut changes.
plus more….

In the end, the leads did a lot of the heavy lifting in reguards to continuity, while the line producers desparately attempted to make tasks for each thing needed for the shots so we would not forget or miss anything!

Paris - Circus Archanus - Night Interior

Sequence Lead: Filip Susteck

The interior circus tent shots were tricky in their own way. Once again continuity was critical and complicated to track. The edit changed around these sequences frequently aswell. The number of shots jumped up and down by the 50s. The creatures were cut out, placed back into the edit, put back on hold…. Originally the sequence was much longer, and was more similar to a real magical circus (if not a sad low budget one as per the plot). But in the end all that spectical took a great deal of screen time, and didn’t offer much in the way of charecter development, so the scene was trimmed WAY down.

Besides the creatures, there was an ENORMOUS amount of roto to be done. The original scene was set, more towards dusk, in the middle of summer, so the tent was brightly lit. Specifically, there was a bright sun bleading through the thick canvas material in the plate photgraphy. When the sequence places in the edit to take place at night….we had to find a way to turn the bright evenly lit plates to dark and atmospheric…..which translated into rotoscoping everything in each shot to layer the light falloff from the center of the tent into darkness at the fringes.

The closeup Kappa shot coming out of a full fluid simulation which was being hit with fire by the offscreen Firedrakes was a last min addition to the film. Very exciting and not at all stressfull for the CG supervisor:)

Portkey

Sequence Lead: Jonathan Turner

This was another one of the ‘DMP’ sequences that Johnathan was overseeing. In this sequence we had to ‘extend’ the plates of Newt and Jacob to have more white-chalky-sandy-grass-hills. Each charecter had to be fully rotoscoped. Eventually we abandoned the idea of ‘exending’ the plates, and just made full background replacements.

The key shot in this sequence was a 1600+ frame beauty of Jake and Newt walking along a green screen of many colors. Camera tracking was super tricky, and then the resulting ‘dmp’ was even trickier. There was no single projection setup in the world which could hold up for that amount of parallax. Eventually, the had to make some CG grass and bushes, and pull some cheats on how ‘close to the ground’ the film camera was:)

Paris - Place Cache

Sequence Lead: Jean-Francois Gagne

The Place Cache was another ‘complicated’ sequence. Not nessasarily because of the type of work needing to be done. The complexity arose from the hord and diversity of unique effects. We saw a climax of different effects and techniques being used to tell the story in this group of shots.

What truely made it tricky to manage and produce, was that no effect, was in more than 2-3 consectutive shots. So each magical effect needed a full run time of development, with no reusabulity. This meant that for every work in progress screening, or change in art direction or edit, the compositing work needed to be re-invented and started from scratch. This was a massive amount of man hours/days to predict and track for production. We spent countless hours re-bidding and evaluating changes in reguard to crewing needs.

In the end, we were pretty happy with the look of the sequence though. I thought all the gold looked pretty when seeing it in the cinema.

Bird Market

Sequence Lead: Jonathan Turner

Yet another sequence where in my minds eye, I see many more shots that we produced for this sequence. Guess the wound up cut out of the film.

Because there were more shots originally, and this was one of the first sequences turned over, we spend a good deal of time look developing the background for these shots. The VFX supervisor had shot loads of nice plate photography for the river and far bank to be used. So we handed all that data off to the environment department, they photoscaned and geo-located the scene, and provided comp with a really nice 2.5D projection setup to re-create the BG in a plausible, phyicially accurate way.

Some of the motivation for do this more complicated proceedure, was the nature of the hero plate photography. The bird market was shot on suprizing long lenses. One shot was a 300mm I think. So they were super compressed spatially. But because the nature and chaos of that set, the amount of detail made the shots ‘feel’ wider than actually filmed. So when we attempting to place BG plates behind the hero scans, the scale always looked and felt off. So we went down the complicated road, to do things corrently first, to validate our thoughts, then cheat if we felt the shots still needed it.

Unfortunatly, all that remains of that work, is 3x shots, two of which are very out of focus! Ha.

Grindenwald Meeting.

Sequence Lead: Jonathan Turner

A small but troublesome sequence to produce. First these were long shots. 700+ frames. Second they were oddly framed and exposed with bluescreen outside the tunnel location. Every time we ‘propperly exposed’ the exterior, the shot looked blown out and un-interesting. When we stopped down he BG, the shot looked ‘cheap’, as if we had pulled some gimic on set to be able to expose for both FG and BG plate. In the end, we shipped along DI mattes with the shots, and they pushed them around in finishing to find the right ballence for the film.

Picket Sewers

Sequence Lead: Jonathan Turner

Originally, there was a whole gimic with Picket. The sequence was much much longer. Picket had a setup where he ran away from some terrifying sewer parasites which crawled out of the water. Then Picket ran along the ground to save the day by picking the lock at the last possible moment. Most of that seems to have been left on the cutting room floor, and what made it to the film was this single cute shot of Picket picking the lock to set our heros free.

Phoenix Chick

Sequence Lead: Steven OConner

These were some of the first shots turned over on the show…at 6k! The studio was keen to get these shots produced asap to test the multi vendor workflow where we pass Minicomps between the studios. Issue was, that schedual for the little bird asset seemed out of sync with the turnover and delivery of these tiny pheonix chick shots. We couldn’t get the shots out of CG:) Anyway. They turned out great, if not a bit last min.

Grindenwald - Queeny

Sequence Lead: Jorge Sanchez Fresno

First off, I rememeber producing many more shots in this sequence. The set extension out the window of Paris (which had to match an exterior being done by another studio…..which turned out to be Framestore London!), floating tea pot gag, and some crew to paint out in the mirrors. Etc, etc.

For reasons I can’t remember anymore, we had touble producing these shots at first. I think it was the prep (paint, roto, keying) which was delayed because of another delivering show in the studio. What I do remember for sure, was Jorge taking charge and leading his team to victory. They made some templates, coordinated a small crew, and knocked the entire sequence out in just a few weeks (once the assets were ready).

But I recall working closer to 25+- shots in this sequence. Not the measly 10 that are in the film…..maybe it just felt like more:)

Family Tree

Sequence Lead: Jonathan Turner

This was one of the trickest sequence of shots to produce!

I see these shots from two points of view at once.

First was living in the moment, and not really understanding how the family tree fit into the story, or what the shots were supposed to look like…..were there even any story boards? No one internally really wanted to touch it (mostly for lack of available time), but also, no department had the tools to complete the shot alone. Animation wanted rigging to make endless new tools. Rigging wanted FX to make some ‘post viz / animatics’. FX wanted some timing, visuals and a clear breif to match from animation…..and comp just wanted some renders comp with.

From the 20/20 vision of hindsight, I see that the cut was changing too much, and the story that the family tree was supposed to tell varied enough from edit to edit, that it made any legitimate progress on the shots difficlt.

That being said. I think they turned out really really nice. No one watching the film would know how close to the deadline these shots were produced or how much trouble we had with them!

Paris - Amphitheatre Battle

Sequence Lead: Steven OConner

The Niffler is easily one of my favourite creatures in the franchise. I don’t know exactly why he’s so cute. He just steals my heart every time on screen.

On the first Fantastic Beasts film, the head of CG condemed the comp team because to make the Niffler’s look (diffuse plus thin film/irridecense) we had to render the creature 3x times and combine the renders in comp. This was not ‘physically accurate’, but more importantly, it was super expensive. Furry creatures are tricky to render. Feathers even more so. Then telling the studio that we need to tripple the render estimates…..it was technique no one was keen to repeat.

Trouble was, that no one could figure out how to make a single feather shader which met all creative requirements:

black looking in all lighting scenarios (not purple, or grey)
irridesence that looks ‘shiney’ but not ‘oily’
all adornment features like the ‘turkey feathers’ on his head always read visually in all lighting scenarios.

I would aquate the problem similar to Thanos in Avengers. How do you make someone look ‘purple’ on an orange planet lit by and warm light source…..because he would in reality turn black or grey at best. We encountered a similar challenge with the niffer. Every scene required a specific shading setup….and so dispite months of lookdev, we rendered same 3x passes most of the time, and sometimes could get away with only rendering 2x. Still very expensive.

Paris - Nurmenguard Castle

Sequence Lead: Steven OConner

These are some more niffler shots that Steven looked after. He was super organized. It was fun and impressive to watch his team work. They were so autonomous. The niffler looked super good dispite only having the minimal amount of time with the final renders to pull the shots off.

Inside the castle was the meat and potatoes of the work in this sequence. These shot puzzles had many pieces to organize and keep track of. Continuity of the BG outside the windows, which shots needs to be ‘deflickered’ because of the changing cut and when the phoenix was present. The cute pheonix chick in his two or three shots. The mountian exploding. The ceiling replacement….is some shots. Despilling some shots, but not others depending on the bounce light….etc. Load of little bits. Each shot varing in techniques, proceedures, and elements needing to be placed in or taken away!