This is how you make Chris Hemsworth dangle off the 57th floor of a Vienna tower
HomeHome > News > This is how you make Chris Hemsworth dangle off the 57th floor of a Vienna tower

This is how you make Chris Hemsworth dangle off the 57th floor of a Vienna tower

Jul 15, 2023

Behind the visual effects work for the ‘Extraction 2’ awning sequence, by Fin Design.

One of the stunning sequences in Sam Hargrave’s Extraction 2 sees the main characters fighting it out on the 57th floor of DC Tower 1 in Vienna.

This action happens outdoors on a glass awning, and inside a gym, leading to several close calls, especially from breaking glass.

The sequence was filmed largely on a bluescreen partial set, with Fin Design then tasked by production visual effects supervisor Bjorn Mayer and visual effects producer Magdalena Wolf to composite in the characters and produce a 360 degree panoramic environment of the surrounding Vienna cityscape.

Here, Fin’s visual effects supervisor Will Towle and CG Supervisor Qazi Hamza Javed tell befores & afters how the intricate CG work, glass sims, comp work and environment build was carried out.

b&a: For the awning area and building (and the weights equipment), tell me about the CG build. What were you basing it on, what kind of scans and photographic textures/reference were captured?

Qazi Hamza Javed (CG supervisor): The gym interior/exterior with all its props and the awning started off as a build based on the scans from the set. Most of the non-reflective props and the building were pretty straightforward. But given that scans don’t pick up transparent surfaces, notably the 3 layers of glass in the awnings case, we did a preliminary build of the glass with what we could make out in the scans and then refined it based on the plates from the shoot. This glass surface had to be a very close match given that all the reflections of the actors were CG and had to match all of the the foot and body contacts.

Light interacting with glass is quite interesting as a physical phenomenon. Deconstructing this behavior from all the references, we knew early on that we’ll need to generate a lot of maps to tackle surface distortion – very specific scratch and smudge maps that built up during the course of the action sequence and anisotropy based on these maps, as the sunlight grazes the surface, birefringence, multiple reflections – across three layers of glass. Once we had a shader and light setup that took care of this behavior, we had the control to artistically direct the shots and push the realism of those surfaces.

b&a: For the wider environment, how was this constructed?

Will Towle (VFX supervisor): The wider environment was a combination of 2D and 3D elements – the production team had done a fantastic job collecting aerial photography at the location of the tower in Vienna and the artists at FIN used this photography to assemble a 360-degree panorama which would serve as the basis of all our environment work. Another benefit of having such good photographic coverage of the area was that – using photogrammetry – we were able to build a simple 3D representation of the local area to serve as a foundation for modeling the main tower and the surrounding structures.

During the 3D build, we implemented a Level of Detail (LoD) system for the buildings & roads in the local area and into the far distance. The roads and buildings in close proximity to the central location were highly detailed, with the level of detail gradually decreasing as we moved further away, ultimately representing distant objects as simple primitive shapes.

To inject some life into the panorama we introduced animated ripples and specular pings to the waterways in 2D and added an overall heat/atmospheric distortion which gave the specular highlights a nice natural undulation, especially when heavily defocused. We also employed Houdini to create a procedural traffic system that simulated the movement of vehicles along roads.

One notable advantage of employing a 3D approach to the environment was the ability to integrate the 3D setup within our lighting scenes. This enabled us to achieve accurate environmental reflections on our hero building, which was an important factor in convincingly presenting it as a physical location.

b&a: What approach did you take to matchmoving, tracking and roto for the live-action elements in the awning fight?

Qazi Hamza Javed (CG supervisor): The matchmove of the digi-doubles had to go through multiple iterations, since we were evaluating them in rendered reflections rather than a CG takeover or for shadow casting. The reflections had to hold up and have the correct contacts from bodies, clothes, boots and fingers. Once we had solid camera tracks and matchmoves, we applied a further refinement pass of shape sculpting in shots to get a very tight match of the deformations that we were seeing in the plate characters.

b&a: How did you tackle cracking and breaking and smashing glass? What were the different ways this was achieved (ie CG, sims, comp)?

Will Towle (VFX supervisor): There were a few different techniques used to help sell the realism of the glass destruction. Firstly, we worked closely with the client-side VFX team to understand their desired behavior for the shattering glass. It was important to lock down what type of glass we were shattering early on, then, depending on the specific scene and the importance of the shatter to the story, we settled on a few different methods.

For wide shots and those quick, blink-and-you’ll-miss-it moments, we opted for a 2D comp approach – a texture revealed at the point of impact – accompanied by a debris hit that consisted of a fine particulate layer combined with slightly larger, chunkier debris – all done in Nuke by our compositing team.

For anything more ‘hero’ we turned to our FX team who did some fantastic shatter simulations using Houdini. Working with both on-set and external references, we matched the physical properties of shattering tempered and non-tempered glass for different situations within the sequence. The setup was designed to offer flexibility for art direction, enabling us to meet the client’s specific requests.

This was particularly evident in the shot where cracks begin to form beneath our characters’ feet, ultimately leading to the whole pane of glass shattering. To achieve this, we combined a rigid body voronoi shatter simulation done in Houdini with 2D textures that are revealed and projected onto the fracturing geometry using Nuke to add extra details.

b&a: In terms of compositing the live action into the awning area, what were some of the very specific compositing challenges that came from dealing with outdoors, sunlight…and glass?

Will Towle (VFX supervisor): The compositing of the sequence presented a significant challenge and I take my hat off to the comp team and our comp supervisor, Adam Paschke. You couldn’t ask for a much more unforgiving sequence, with a rapidly moving camera, dynamic action, a bluescreen requiring replacement with a bright background and the need to properly handle reflections and refractions with accurate depth cues and defocus.

As for the challenges associated with the outdoor setting, fortunately the set was constructed outside and primarily lit by natural sunlight. As a result, we rarely needed to adjust the lighting in the footage. Early discussions with the client led us to always align our work with the direction of light in the original plate, even if it varied slightly from shot to shot. This ensured that we never worked against the photography.

One additional obstacle we faced was the changing white balance of the shots due to the shoot spanning several weeks. To address this, we implemented a neutral grading workflow. All plates were neutrally graded and composited with our neutrally graded CG, minimizing the need for per-shot grading and maintaining consistency throughout the sequence.

The glass elements posed both technical and artistic challenges. As mentioned earlier, our 3D environment setup was integrated into the lighting of the scene, resulting in accurate reflections and refractions in our CG renders. To maintain flexibility in our pipeline, we devised a workflow to separate the reflections and refractions in comp, allowing us to defocus and depth-cue them independently from the hard-surface objects like the glass hardware and support beams.

On-top of this, all the reflections of the characters in the sequence were created digitally. These were modeled, look-devved, and lit in Maya/Houdini and then rendered with our beauty renders of the awning. Mattes and AOVs were provided to comp to tweak the lighting and depth-cueing of all the CG components, ensuring they matched the plate as closely as possible.

b&a: For the awning area and building (and the weights equipment), tell me about the CG build. What were you basing it on, what kind of scans and photographic textures/reference were captured?Qazi Hamza Javed (CG supervisor)b&a: For the wider environment, how was this constructed?Will Towle (VFX supervisor)b&a: What approach did you take to matchmoving, tracking and roto for the live-action elements in the awning fight?Qazi Hamza Javed (CG supervisor)b&a: How did you tackle cracking and breaking and smashing glass? What were the different ways this was achieved (ie CG, sims, comp)?Will Towle (VFX supervisor)b&a: In terms of compositing the live action into the awning area, what were some of the very specific compositing challenges that came from dealing with outdoors, sunlight…and glass?Will Towle (VFX supervisor)