Category: Advanced & Experimental

For this project, we were divided into teams which included different roles. I claimed the role of CG lead for team Y, whose chosen theme was Lo-fi. This genre of design is best compared to the style of Alien (1979) and the original Star Wars (1977-) movies. With this style in mind, we went more specific and decided upon the style of ‘cassette futurism’. This included geometric computer design, VHS tapes and game cartridges, and generally centres around the idea of a futuristic 1980’s aesthetic .

Our team members are as follows:

Look Dev Artist: Tim Schweizer
Look Dev Artist: Kaiyue Wang
Look Dev Artist: Tianqixu
Look Dev Artist: Jess Howard
Look Dev Artist: Margita
Look Dev Artist: Ruijia Zhang
Look Dev Artist: Dominic O’Donnell
Look Dev Artist: Saurabh

CG Generalist: YaFang Zheng
CG Lead: Daniel Loutfi

Compositor: Guoxin Lin
Compositing Lead: Mingyue

Real World Objects

We first set out to collect and gather resources to fill our mood board and asset list. I saved several images which followed the art style and genre, and also several references to real-world objects which fit this aesthetic.

Weekly Development

Asset Creation

In order to kick-start my involvement in this collaborative project, I first wanted to attempt modeling in our chosen style to get a feel for the concept. As cassette futurism tends to include fairly geometric shapes, to Tim Schweizer’s credit this was a very achievable art style in terms of 3D modelling and maintaining consistency.

To this regard I wanted to model the key features of tech in cassette futurism, namely: thick ribbon cables, rotational dials, ports and cutouts, numerous sliders, and small greenlit screens. I attempted to maintain a realistic bevel amount for the plastic edges to get a more molded shape. For the helix antenna, I thought this might be more interesting than a simple plastic wand, and I created this in ZBrush. These assets could be used as filler objects for the scene if needed, as there are several tech areas of the scene.

As my role requires that we maintain a consistent style amongst several 3D asset developers, having a dynamic 3D reference towards shapes and features can be very helpful.

Week 2 – 17 April

Pre-Vis & Layout

In order to get the room dimensions & camera perspective, I took a screen shot of the empty room, and loaded the image into fSpy. I had utilized this software in my BA Animation course for match-move purposes, and it was very quick and effective at getting camera parameters. I then imported the fSpy file into Blender, changed the coordinates to match Maya’s orientation, and exported the .fbx files into Maya for pre-vis.

For a consistent sense of scale, I brought the Unreal mannequin into Maya and reduced the topology so it would be light for our group to import into their scenes and use as a measurement aid. I linked the model into our Google Drive folder for easy access by the team.

For the last section of pre-vis in Maya, I created several low-poly placeholder assets to get an idea of the layout for our team. I used an HDRI for the exterior, and a light portal to minimize noise. I used several modelling tools to develop this basic outline of the perspective and layout.

I also wanted to test my models by importing them into the scene.

Overall, these techniques seemed extremely effective and quick at granting the development pipeline a strong idea of what the final project could look like, and there’s a very good chance I will incorporate these techniques into my FMP.

Week 3 – Monday 24th April

Now that we have received the footage, our compositors have completed tracking the scene in 3DEqualizer. I continued to keep advising and trying to help our 3D asset developers as much as possible by suggesting certain changes to aspects of their models to fit our scene, fixing topology, and trying to generally be as useful as possible, whilst staying in communication with our composition team.

Un-distorting the Plate for Maya:

Get Distortion Values (D: 0.07 QD: -0.04)

I also began tracking the scene in 3DE as this was our next task for the upcoming Friday. There were some asset changes, as the projector was quite a difficult shape to manage, so we switched some assets assignments around and I decided to take on modelling the projector.

Team Asset Development & Quality Assurance
(Applies to Entire Length of Project)

As some models are starting to make their way through the pipeline, I had a look at each model and in certain areas tried to suggest some ideas concerning topology and texture mapping. I had a strong feeling that UV mapping and texturing assets were going to be one of the more challenging aspects.

Margita Alien Tube

Dominic Glass Table

I assisted Dominic with his table and sofa models, and we made several quality-of-life changes to certain areas of his assets, with UV’s and topology. We ended up patching over the star-shaped hole in the table, as this was creating more issues than it needed to, with stray vertices and uneven subdivisions. As I showed Dominic what was achievable with texturing, even as a very rough example, we moved ahead with the changes. Once this was bridged over, and the topology corrected, we both decided that texturing would have a much greater, and more successful design to the final asset.

Furthermore, I kept a close eye on this asset and was not too happy with the table’s legs. To this effect, I sat with Dominic to help him create something less angular, and more visually appealing. I suggested two options: creating legs with an ‘X’ shape, with nicely beveled edges, or alternatively, simple rounded and tapered cylinders. His decision was the former, and we worked to develop this asset.

Dominic Couch Asset

Whilst doing these high poly bakes and texture tests were not strictly my role, I still wanted to envision what certain assets in the scene may look like once completed. As someone who personally enjoys pushing models to their completion, I could not help but attempt a quick texture test. This was also helpful by granting me further insight into how the model was created, and how the UV maps could potentially be laid out.

I also wanted to assist our look dev artist further with areas of the models that could be improved, and more importantly, to garner more interest in the final look of the model by reaching as many milestones of the 3D asset development pipeline as possible.

I sat with Dominic for several hours whilst assisting him with developing his models. We went through as many aspects of 3D modeling as I could possibly explain given the time. Namely we went through: modeling and correcting topology, UV mapping and layout, texel density and resolution, assigning materials and shaders, general mesh cleanup (Deleting history, freezing transforms, fixing N-gons), and a short rundown of subdivision surfaces.

The next objective in the pipeline would be texturing, which in this case I would like Dominic to keep somewhat simplistic, whilst also visually interesting.

Tim Cactus Asset

Our look dev artist Tim uploaded his cactus model to the drive. Having seen the asset and parts of its development in class, the look was very good and it was obvious that a lot of work had been done to develop the model. However, there were several optimizations that needed to be made to bring the asset into the scene.

First off I noticed the file size was quite large, indicating a high polycount. In my eyes, a 3D asset can be considered ‘heavy’ if the file size exceeds 25MB, and in this case the size was 50MB so it was clear that optimization needed to be done. Opening the model in ZBrush confirmed this, with a polycount of 1m+.

I proceeded to do a quick texture bake on the model, to see if we could replicate the spikes somehow with normal and height maps. Personally I would say this is the way to go with this asset, and use colour ID maps to texture with ease. I gave the model some UV’s and materials, and began some tests in Substance Painter. I did not complete the texturing as this is not my role for this asset, and used my work as an example to Tim who I contacted about the asset, and what we can do to make it usable in the final scene.

Saurabh Floor Asset

There were some major issues which I encountered when importing this particular model into our Maya scene. The worst case being a specific piece of the model which was breaking Arnold, and causing the whole scene to appear black. As I had placed this scene into our Drive for our CG generalist Yafang, she straight away told me that the render was not working at all. I had a closer look and found the broken piece of geometry, which was a center panel created with Booleans. As a temporary workaround, we removed the panel from the scene, and the render was able to take place.

As I had been notified by the artist that the development had incorporated Boolean geometry, I predicted earlier that this was going to create issues in rendering and topology if not looked over carefully. My prediction however paled to the reality of the final asset. Despite receiving approval in ftrack, this model was clearly not ready and require much attention to be usable.

Week 4 – Monday 1st May

New Pre-Vis for Working Plate

For this week, I re-created my pre-vis of the scene with the camera track and locators exported from 3DE. I exported the models from my original pre-vis to illustrate layout in the final plate. I also created a drone model as a placeholder for our look-dev drone modeler, and animated the model to fly around the scene with the camera.

As I was asking some of our look dev artists about their assets, I discovered that some of the team were working on effects such as small explosions which are detonated by the drone. It was at this point I suggested that instead of the drone existing as a piece of weaponry, the drone could instead be a guide, or a companion, which shows the camera controller around the scene. Also, the drone would have a beam of holographic light which illuminates areas of interest, and leaves behind a smoke trail. The team liked the idea, and we moved forwards to develop this further.

Another obstacle however lay ahead, as the grating in the floor was making the water feature very difficult to see. As we wanted to have animated elements such as fish, or eels swimming in the liquid, seeing them through the heavy grating would not work very well. I then set out to find a better location, and decided upon placing this on the right wall, with an area light set behind the “glass”. I showed this to our team and they also decided this was a better placement for the liquid feature, with a much clearer view of the animations.

After speaking with Gonzalo about my next steps, he asked for a top-down screenshot of all the assets and layout. He also asked for a lighting test to match the original footage. To create this lighting setup, I used photometric Arnold lights with an IES profile (freely obtained from IES-library.org) in an attempt to match the shape of the light in the footage.

Originally, as I had gone ahead and done the pre-visualization and worked on the layout early, our CG Generalist and myself decided to switch tasks, where they would be doing the lighting, and I would focus on layout.

Week 5 – Monday, May 8th

For this week, I continued to oversee asset development for our team, and sent several reference images to our chat to ask which texture style people liked the most. As we now had UV layouts for the first models, creating a consistent artistic blend between all the models and their textures would be important. The image below is what was chosen by our group, as they felt it was the most fitting to our aesthetic. I also agreed and continued to find more reference images similar to the example.

I also wanted to create some trim-sheets, as I had learned in my BA that these are an effective way to provide environment artists with several different panels which maintain the style, granting the artist more flexibility to develop the scene further. This technique is mostly relevant to level design in game development, but can also be powerful for a task such as this.

Week 6 – Monday, May 15th

At this point our team have some assets ready for importing into the scene, and I started to rebuild the environment in Maya to Mingyue’s track. She had received feedback from Dominic that the track was accurate. I kept in communication with as many of our look-dev’s and compositors as possible, providing consistent updates in our WhatsApp group and on our Padlet, which I also asked everyone to try and keep up-to-date.

As Gonzalo had asked for a lighting and render test, we rendered my pre-visualized scene whilst I was rebuilding the scene according to Mingyue’s track. It was for this reason that the track ends earlier than the complete timeline of the footage.

As well as organizing the matchmove and importing the scene, I moved to develop the ceiling of the footage to make it more interesting. This was one of the assets that were originally assigned to me, so I began to work on the models.

Week 8 – Monday 22nd May

At this point we moved to incorporate the lighting into the scene. We also added the screen in the top left to display some type of technical animation as a UI element, and I heavily shifted the assets in an attempt to create more balance. As our drone animator was using my animation as reference, I also changed the animation to make it more ‘playful’ for her to follow. I removed my projector asset because it no longer fit the atmosphere very well, and we instead opted to create a projected animation of a spinning planet. I also imported Tim’s cactus into the level.

As Gonzalo had told us that each look dev artist had to model three assets each, we further developed our asset tracking sheet to reflect what had to be done.

Week 9 – Monday 29th May

For this week the goal was to render out a “Master” version of the scene, with the first assets from the team replacing the original placeholders.

Throughout the week I kept a close eye on the animation of the drone, as we wanted to comp in some smoke particles which surround and follow the drone as it moves through the scene. Also I’m checking every asset that is uploaded to the Google Drive to check polycounts, UV’s, and textures.

Week 10 – Monday 5th June

This week I was getting a little desperate for the finalized assets and textures from some of the group. The floor especially was not seeming to get anywhere despite repeated requests. Due to this, and our groups need to allow the compositors a good 2 weeks to work on the scene, I moved to model a backup floor asset and textures in the off-chance that this ends up being one of the assets everyone will be waiting for.

I also continued to assist with other look-dev assets and textures, mainly for the couch and table.

Week 11 – Monday 12th June

For this week our look dev artist responsible for the floor asset returned, and we sat together for several hours while I taught him to use Substance Painter, how to correctly set up UV’s, and finalizing the model. This is something I’ve done several times at this point to help certain members of our development team progress.

Myself and others from our team joined the meeting with Dominic, and took note of his advice, as well as the tasks he required.

These were:

Creating a Pinterest mood-board
Fixing the Lighting to emanate from a single source. (Light Portals + HDRI)
Completing an animated rendered sequence.
Creating Invisible Planes (Light Blockers)
Importing the Aquarium assets.
Fixing the Aquarium Lighting Issue (Create inner Faces)

Right now the pressure is rapidly climbing, due to not having certain areas of the scene completed yet with assets and textures.

The amount of troubleshooting required for this project was a fair amount, as we encountered several major issues and crashes. This caused some frustration when it did happen, but so far we have been able to find ways to fix these problems when they do happen.

I did some test renders of the new lighting, and immediately noticed that the scene was very dark. I took the renders into Photoshop and boosted areas which needed to be brighter to be shown to our lighting specialist. I also wanted to tell the compositors to add the low hanging fog over the vents.

We needed a wall for the right-hand side of the room, where the aquarium is positioned. I created a low-poly base and exported into ZBrush to create a high-poly details and normal maps.

To further my skills in Substance Painter, I researched how to use Anchor Points, Micro-Normals, and Micro-Height to give textured normals a curvature map for smart masks to reference. These were extremely effective.

Critical Reflection

This project was quite a challenge to complete. There were many things to consider, with balancing the layout being the most difficult. Communication with the team was very strong, and I feel that my suggestions were well heard, and ideas integrated into the scene. I attempted to consider everyone’s input when it came to layout, composition, and any other aspect. The working CG render of the final sequence was thwart with issues. From disappearing textures, to broken geometry, and lighting issues, it was truly a challenge to get a successful render in place, of which I have to pay my regards to the composition team for working through these problems.

. As I had researched the CG lead position, a main factor of this role is to assist in the development pipeline which I truly believe I achieved, having assisted many of our look dev artists in modeling, clean up, and texturing. I may have sat for over 36 hours in total when in comes to teaching UV mapping, Substance Painter, implementing Bridge materials, and correcting topology, which I was more than happy to do so. I had to make some suggestions when it came to polycounts and textures, and these suggestions were well taken by the team. Our look dev artists did an excellent job at updating me with their most recent models and textures, and notifying me of time they would be away.

When it comes to the overall timeline of the project, I feel that we started off very strong, with the pre-vis in place, there was some idea of the general layout of the room. I would consider this to be an important step towards starting the development cycle. Once we had assets making their way through the pipeline. the scene started to take place, and we maintained a consistent flow of building up the environment. There were some worrisome times, where certain models were taking too long to complete, or the models having broken geometry, but once these issues were communicated and fixed, as well as the aquarium asset being one of the most difficult to implement, the workflow was resumed and we maintained the schedule.

After solving all of these issues, we had a successful render in place. From a personal view, implementing all the different models and textures was quite a challenge, especially when it came to naming conventions. I made my best effort to develop the scene and showcase every persons contribution, this was quite difficult considering the amount of look dev artists. These challenges were overcome for the most part, and the project was finalized. I feel that we matched the tone of ‘low-fi’, and we also followed the objectives we set to achieve,

Final Music Video

Artist : 5endi

Title : No Hablo Espanol

Editors : Domanic O’Donnell – Robin Carreras – Daniel Loutfi