I assume we all have seen and enjoyed “The Witcher” on Netflix – and most of us have seen the cinematics for the games, and a few have even played them. Reason enough to pull it all together and talk to the guys who made the cinematics and quite a few shots for the show: Platige Image.

We talked to Visual Effects Supervisor Mateusz Tokarz (whose credits include projects like “Safe Inside”, – “Lilyhammer” and “The Giant Mechanical Man”), compositor and Art Director Rafal Sadowy, whose recent projects were for example “Wonder Woman”, “Another Day of Life” and “Antichrist” (still a favourite of the editorial team), and VFX Producer Krzysztof Krok, who worked on “Lilyhamer” as well.

DP: The obvious first question: Did you get to play “The Witcher” as official prep for doing the VFX on the show?

Mateusz Tokarz: Lauren Schmidt Hissrich, the showrunner, stated clearly from the very first day that our world is going to be based on books and stories, not video games.

DP: Since you did the cinematic for “Witcher 2”, was it fun to get back into that world?

Mateusz Tokarz: It’s great to be part of both. Game cinematics (there was more than one!) for “The Witcher” games were ground-breaking and after all those years still look great. They also started a new era in the history of Platige Image – the era of game cinematics, which became an important branch of the company. Now we’re part of the biggest Netflix show of the year, showing that we are reliable partners for film and TV as well. “The Witcher 2” game was a long time ago, but for people who worked on any of these cinematics, it was definitely fun to get back to this world.

Rafal Sadowy: For me, it has come full circle, because many years ago, I was involved in one of “The Witcher” cinematics – it was my first great opportunity to work with Tomek Baginski – their director at that time. It was great to come back to this subject and look at it from a different perspective.

DP: How many shots did Platige work on? And what was the briefing for those?

Mateusz Tokarz: It was around 150 shots. We cooperated closely with Julian Parry, the overall VFX supervisor. Our journey started at the end of last year (2018). We were helping production with some RnD before the shoot. Then we were one of five studios working on VFX for the first season. We were mostly responsible for magic effects – portals, spells, signs – but also environments and a lot of other various types of work. Our main purpose was to create effects that are subtle and elegant. We’ve spent a lot of time researching practical physical phenomena to be able to create something interesting and unique. In this case it wasn’t easy – there are not many real-life examples for magic.

DP: How many people did you gather in your team to make the shots?

Krzysztof Krok: The team was quite big – approximately 130 people involved in total, but the core team was around 30.

Mateusz Tokarz: We had a lot of hardcore Sapkowski fans in the studio who wanted to contribute in some way, just to be part of this saga – we were super excited about this project, but it was very intense, and this gave us extra power.

DP: What was your personal favourite shot to work on?

Krzysztof Krok: One of them was Yennefer’s and Istredd’s portal created in Aretuza (2nd episode). It was one of the first magic effects that we started developing, very early in the production. It evolved and needed some adjustments when we got the raw footage from the set. We understood then how flexible the VFX setups have to be to make quick adjustments as well as our thinking on how magic has to be.

Mateusz Tokarz: My favourite sequence is one with the Shan Keyan Tree in the enchanted forest. We had to show it in shots during the day and in Ciri’s vision as well. It has very vibrant color correction in the trailers and fans loved it. On the other hand, in some shots from this sequence we replaced around 75 % of the image with CG renders and it was demanding to match the very stylised lenses that were used there.

There’s also a funny story about the making of this episode, which came with one of the first crucial deadlines we had to meet. On the day before, we were working long hours and agreed we’d come in earlier the next morning since it was the deadline day. I came a bit late and realized that Rafi, our Art Director, hadn’t come at all.

Chris texted him and after a couple of minutes we learned Rafi wouldn’t come as he was hospitalized. I just didn’t register it, this couldn’t be true, so I said: “Alright, call me when he comes,” and got to work. After 15-20 minutes I asked again what’s going on, and Chris showed me Rafi’s photo in a hospital bed. He had broken his arm riding an electric scooter.

It was one of these moments when you’re doing crazy overtime, but you’re so pumped up that it’s hard to believe that something may go wrong. Fortunately, these sequences look fabulous and maybe because of that I really like them.

We’ve also taken care of many shots in the final battle. The entire crew did a wonderful job on set and thanks to that it was a pleasure to contribute and make it even cooler.

Rafal Sadowy: All shots that include different kinds of magic are my favourites. Mainly because they were the most challenging and the most interesting to figure out. For example, for the portals, it was mixing different physical phenomena together – liquid surfaces and spiral clouds often seen in satellite imagery.

DP: Can you walk us through the creation of the portals?

Mateusz Tokarz: We were developing them before the shooting started. Our goal was to create something subtle and unique. There were tons of static concepts that we did before we started FX tests in Houdini. In terms of shape, our inspiration was galaxies and nebulas. In terms of appearance, physical effects that involve light aberrations and dispersions.

DP: Can you walk us through your pipeline for the show?

Mateusz Tokarz: We’re using pretty much everything that is industry standard these days. Our main 3D software are Maya and Houdini. Substance Painter for texturing. Nuke for compositing. PFTrack for camera matchmove. Everything is connected with the use of our in-house pipeline that we’ve developed over the years.

DP: Any particular new tools you used?

Mateusz Tokarz: We’ve created a few smart tools to automate turnovers and deliveries. That helped us to keep everything in order during peaks of work.

DP: With you being known for your work on cinematics and a few films and shows under your belt, what are the different forms of approach you have adopted, and which parts were easily transferable?

Mateusz Tokarz: The difference between feature films, cinematics, commercials and TV shows is the speed of work and the frequency of client’s feedback. TV shows are definitely the most intense – there’s almost instant client feedback and many levels of approvals.

Krzysztof Krok: The main difference is that you are one of the vendors. Part of something bigger and greater. You have to keep the pace. You have to adapt and at the end of the day understand there are many more teams involved. Your regularity helps everyone around and a daily routine is highly desirable.

DP: With you, Framestore, and Cinesite sharing the VFX-load, how did you choose who did which shot and how was your teamwork for that?

Mateusz Tokarz: Everything was managed by the Netflix post-production team. There were a few shared shots between us and other studios. In most cases, both studios were working simultaneously, but always one was chosen as the finishing studio and responsible for delivering complete shots. When the other studio’s part was finished, they would send their final comp render and then we would apply our effects on top of this. For a few shots we exchanged FX setups, DMPs and Nuke scripts.

DP: Some effects had quite a bit of practical FX involved: Did that make it easier for you? (At least some things were shown online.)

Mateusz Tokarz: It’s always easier to enhance something rather than create from scratch. I’m a huge fan of good balance between practical effects and CG involved in one shot. These types of shots always look very good.

Rafal Sadowy: The post-production process always becomes smoother when we use practical elements on the set. Even if we are replacing them completely with CG elements in the end, they are the best reference to integrate CG with live action plates.

DP: If you had to start that season all over again, what would you do differently?

Mateusz Tokarz: We had to learn during the work what Julian’s, Lauren’s and the producer’s expectations are. I think that it’s not about doing things differently.

Rafal Sadowy: Not riding electric scooters!

InstaLOD automates the entire 3D workflow, including the tedious steps that no 3D artist likes to perform, such as UV unwrapping, manual retopo or baking. You can see an overview of InstaLOD’s complete feature set in DP 03:19. In this issue, we take a closer look at the new features and functionalities that have been added in the latest update to completely revolutionise enterprise workflows.

In December 2019, InstaLOD released a huge update in which over 500 improvements and new features were added to their standalone software InstaLOD Studio XL.

Among other things, these features speed up work within the software and make many steps of data preparation more intuitive than before. In addition to these many quality of life improvements, InstaLOD has launched several brand new products that are set to completely change the way 3D CAD and polygon data is processed in the enterprise market.

AMD ProRender

Let’s start with a feature that is a welcome addition for any 3D hobbyist, freelancer or enterprise customer: easy to set up final shot rendering.

To prepare PBR models for product renderings by default, you have to be prepared to convert all materials to be compatible with the respective renderers, which in short means: every material has to be set up from scratch.

Render engines such as Arnold, Redshift or V-Ray have their own proprietary shaders, so you can’t simply use a Lambert shader to imitate complex materials such as clearcoat, subsurface scattering or sheen. InstaLOD automates what typically has to be converted manually with the integration of AMD ProRender. The PBR material is automatically converted into a ProRender shader.

This means you remain flexible in the PBR model and can simply run ProRender at render time and sit back – without having to manually re-map textures.

The Material Editor has also been improved; shading balls have been added for quick previews of materials. Adding and filtering materials is also quick to do, and the editor is easy to understand at first glance, with value sliders and texture input options per material parameter for easy editing.

In addition to the standard PBR parameters (Colour, Roughness and Metalness), the Material Editor contains many new parameters that can be used for raytrace rendering in Radeon ProRender. These include, for example, refraction, sheen, clearcoat and SSS. This means that high-end renderings can be set up with just a few mouse clicks, as shown above.

The integrated post-processing stack enables users to edit rendered images quickly and easily. Optional image processing functions such as colour correction, exposure, depth-of-field or AI denoising can thus be carried out locally within InstaLOD without having to open external image processing programs.

InstaLOD Grid

In companies, it often happens that employees cannot fully utilise certain applications. Application-specific training must then be provided or, alternatively, colleagues from other departments must take over the tasks.

InstaLOD now wants to change this with its new product InstaLOD Grid. InstaLOD Grid makes it easy to set up a local InstaLOD computing cluster within an organisation. This enables employees to access InstaLOD Grid via an InstaLOD Grid web application. Within the web application, users have the option of uploading their data and editing it using ready-made profiles.

Profiles are added by InstaLOD experts within the company when the master server is set up, which is quick and easy to do.

Profiles are saved as JSON files by default, which can easily become confusing for beginners in their text form. Therefore, this was simplified for the web application so that the user does not need to know the individual operation settings, but only a title and a description are visible in the UI, so that everyone knows that profile XYZ, for example, creates an LOD chain without having to understand any settings within InstaLOD. This means that employees do not have to attend application-specific training courses to be able to fully utilise the InstaLOD functionalities and settings throughout the company.

When a user creates a task, it is received by the InstaLOD grid server and sent to an available processing slave for further processing. Once the data has been processed, it is automatically sent to the server where users can download the data. To summarise, you upload data, specify a profile preset to be used for processing, let InstaLOD perform the operations and download the finished data at the end. For comprehensive system integrations, InstaLOD Grid can be accessed via an integrated web service that can be connected to other web services, PLM systems and conventional data pipelines. This would make it possible, for example, to load the models of 3D shops directly into their libraries after processing.

InstaLOD WatchDog

Compared to VFX or game studios, large companies often have to process tens or hundreds of thousands of data for visualisations and other purposes. Due to their PLM (Product Lifecycle Management) systems, the data often changes weekly, if not daily. As a result, it quickly becomes confusing as to which data is new or has already been prepared, and processing such masses of data manually is almost impossible.

The process of data preparation is therefore not scalable in the traditional way, which is why InstaLOD has entered the market with a new solution. InstaLOD WatchDog is a powerful tool that can process and correctly mirror complex folder structures. After the initial processing of data, it remembers which files have already been processed so that data is only processed if it has been changed or newly added. This prevents problem areas where thousands of data would otherwise have to be completely reprocessed in the event of the smallest changes.

By combining InstaLOD WatchDog with InstaLOD Grid, you can create a fully automated ecosystem that makes it easy to scale data processing. Changed data is automatically recognised, sent to InstaLOD Grid, processed and then stored in the correct folder structure. To simplify the system even further and to avoid having to trigger InstaLOD WatchDog manually when data changes, the InstaLOD Scheduler can be used to specify a specific schedule using cron expressions, specifying the intervals at which InstaLOD should check and process the data for changes or new data. This allows the data preparation system to run completely automatically without the need to initiate manual steps.

Scene Rules

We looked at InstaLOD’s Scene Rules in the article in issue 07:19 (presumably meaning DP 06 : 2019 or DP 01 : 2019). DP 06 : 2019 or DP 01 : 2020. Issue 07 : 2019 does not exist. Editor’s note). The issue explained how this set of rules can be used to perform complex actions when importing data with a simple setup.

This includes actions such as adding materials, deleting objects or selectively tessellating certain CAD parts. These scene rules have also been revised in the 2020 update and expanded with many new predicates and possible attributes so that users now have even more flexibility when processing. Actions such as “Delete all screws that are smaller than 10 cm” or “Close all holes on objects that have more than 3 boundary vertices” are now a simple task.

A new Record option has been added to the Scene Rules, allowing users to automatically record rules without having to set them up manually. You simply press the record button and set up the scene. Every change is automatically saved in a rule. The recorded profile can be saved and recalled at a later time so that the same steps can be carried out fully automatically.

Mesh Match Rules

Since the December update, the above mentioned Scene Rules are now also available within the Mesh Match Rules. This means that you can control which objects are processed by which mesh operations. This option was already available before, but the assignment of objects to operations was limited to the names of the objects. This has now been extended with the complete arrangement of attributes, which means that the size, position and materials of the objects can now be used to trigger operations.

Final Words

As you can see, InstaLOD has taken an already powerful tool and turned it into a real high performer that knows when to process which data and where it belongs – a fully automated ecosystem for data preparation that can be completed with product renderings for visualisations.

More information about InstaLOD can be found on the InstaLOD.com website or on InstaLOD’s YouTube channel.