Modeling/Sculpting Stage

The first stage in my pipeline is the Modelling and sculpting stage. My initial research tested some modelling techniques, but as per my feedback, I looked at modelling a small asset to determine the order in which I would geometry model, and sculpt.

Because I chose, also, to reevaluate the Software I used for Sculpting, I have tested a few assets, rather than just one asset, to get an idea for the advantages and disadvantages of sculpting certain things. For example, Hard Surface Modelling in 3DS Max can require a lot of polys to avoid the default "sharp" edges, but is generally easier to accomplish compared to sculpting flat surfaces in ZBrush.

Softer Edges in 3DS Max using Turbosmooth

Kukri Knife modelled in 3DS Max to test Hard Surface Modelling Techniques 

Sculpt of Mountain in ZBrush, ready for retopology

Pillar 3D Modelled in 3DS Max, then sculpted in ZBrush, and textured and baked in Substance Painter

My tests aimed to:

• Determine which methods in the modelling stage would be the most efficient
• Which methods would also allow me to produce a quality of work that is portfolio-ready. 

The Kukri Knife took significantly longer to model in 3DS Max, but produced a model that needed no retopology and came in at a lower poly count than any of the ZBrush sculpts I produced.

The mountain used ZBrush entirely to model, handling retopology after the sculpting process. I found this process the most time consuming, but could understand the necessity of it for sculpting organic modelling in a short amount of time.

The method used for the pillar involved geometry modelling in 3DS Max, then sculpting onto this primitive mesh in ZBrush. The high poly model was then baked onto the original, low poly mesh using Substance Painter, reducing the poly count to under 1000 polys.

This method ensured that the geometry was never dramatically warped during the sculpting stage, but it would depend on the complexity of the geometry being produced in the final product.

Next Step

These tests helped me to determine a workflow for this stage of the pipeline, so I would, therefore, like to stress test it. I have found that sculpting from scratch in ZBrush is not advisable with my current level of skill, and so these further tests will focus primarily on the Geometry modeling (3DS Max ) >Sculpting (ZBrush) > Baking/Retopology (Substance Painter)

There are also different methods of retopology, using different software, which I feel i can test in more depth, dedicating a post to just the findings from that.

Sculpting Program

By undertaking this testing phase again, it gave me a chance to further investigate which of the two sculpting programs to use. Whereas I previously only had access to Autodesk Mudbox, now I had the chance to use ZBrush and compare the two.

Mudbox definitely used a more familiar layout that I was accustomed to, being part of the Autodesk Software. This meant there were a lot of shared traits with 3DS Max, and I had also learned my way around the software during the first few weeks of my project.

However, while ZBrush lacked this familiarity, it made up for this in the intricacy that it allowed one to sculpt. An array of brushes were available with precise settings for each brush, and any deformations and modifications one wished to make on the model.

Setting up Reference images was intuitive, integrated into the scene so that the sculpting was easily comparable to the source images at any point during the sculpt. This is in contrast to the limited projection settings provided by Mudbox.

In terms of learning this new software, ZBrush was the more popular among 3D Modelling forums and websites. The tutorials available were more in depth and plentiful, especially on Lynda.com, my preferred means of tutorials.

Next Step

The test that I want to undertake makes use of the sculpting software, so after learning it, I would like like test some assets out in it to determine this area of the workflow I am wanting to follow. By doing this, I hope to assess whether it is preferable to sculpt from scratch in ZBrush, then retopologise, or to create a poly model in 3DS Max, then sculpt in ZBrush and return to 3DS Max to retopologise. 

The method of retopology also needs addressing, as many programs are capable of doing it, but I need to figure out which is the most efficient, effective and professionally used, if applicable.

Sources

Munoz, P. (2018). Learning ZBrushCore. [online] Lynda.com - from LinkedIn. Available at: https://www.lynda.com/ZBrush-tutorials/Welcome/648907/690587-4.html?autoplay=true [Accessed 1 Feb. 2018].

Feedback

Having received my feedback on my Project proposal, I have had to change some elements of my milestones, though there aren't any major setbacks.

Firstly, I may need to reconsider my final artifact, which I had felt may be over-scoped. Whereas my final artifact idea makes use of the hard surface modelling techniques, there is very little room for any organic modelling, which is cited as a requirement in a lot of job advertisements and so I would like to display this in my portfolio.

My feedback details a process I had not documented, using re topology tools and so I plan to use an asset I have previously modelled to take it through this pipeline as suggested. Any models used in the tests can be documented in my portfolio if they are to a high enough standard and, as suggested, could define what the final product would be, and its limitations a bit better.

Although I had considered using high poly sculpting to low poly sculpting, I thought I may be able to bake to a lower poly mesh, but had not accounted for any new geometry said sculpting creates, so this is a valuable method to be reiterating.

Modelling Update

I've been looking at building an environment using Modular assets.

My first iteration of this was to build the modular assets so that they could be used to construct buildings for a town, or small city. The kind of environment seen in towns with pedestrian roads. I researched the different elements of buildings, and worked on splitting the buildings into ground floor models, and upper stories, so that a level designer could construct them at their ease to produce a wide variety of types, while still maintaining a consistent style.

The tests I produced demonstrate the ways that the buildings could be split up to be modular. This also provides me with a sense of the scale of the project. For example, how many polys the individual assets will be and how much they will amount to. This data was taken from sources

Modular Roofs. I researched into different types, and worked on fitting them together

Very primitive examples of working out the pieces I could produce for the building

Storefront that uses tileable ground floor and windows

Experiment with using textures to pull out geometry for the model

Style Guide

A collection of images that I have gathered to hone down the style of building that I am hoping to produce. These images give me a good understanding of the different parts of the building I can split up and consider for the modular approach. For example, I could produce a variety of different cornices with different patterns and dentil placements which could seamlessly attach to a variety of building facades.

I have also researched into the specific parts of buildings, which will make it easier to search for specific photo reference and refer to conventions that actual buildings follow too.

Source: Nevin (2007)

For aesthetics, I originally planned to theme it on the streets of Akihabara in Japan, which 'Cyberpunk' draws a lot of reference from. Using bright neon lights against a dark, blue toned backdrop for contrast. Where the signs on the side of buildings for advertisement were so plentiful, it seems like any available space is used for the signage. However, I wasn't too sure about if this would be possible in the time I had, having to model these assets, and devise a system of ensuring they fit together in a modular structure.

A collection of images highlighting the palette and recurring elements of 'cyberpunk' cities

Images detailing a brassier sci-fi interior, showing more recurring elements of the genre. Hard surfaces, bright lights and rounder geometry than contemporary architecture. 

Sources

Nevin, S. (2007). Building Parts Diagram. [online] Gallery.insanart.com. Available at: http://gallery.insanart.com/v/design/buildingparts.jpg.html [Accessed 23 Nov. 2017].

Poly Count for buildings

While trying to determine the poly count for the project I am undertaking, I considered how the entire project would need to be optimised to run in engine. The same thing would need to be considered if working in a AAA studio, and is cited as a requirement in most of the job advertisments I looked at.

Having decided on making a building using PBR texturing methods and utilising Polygon-sculpting methods, I looked at similar games. Considering I had looked at job adverts from Rockstar, I researched into the polygon counts seen in games like Grand Theft Auto V, Infamous Second Son. Games that use cities as their environment.

Another source I found is the Unreal Marketplace which, as the engine I am wanting to render the assets in, is useful in providing professional constraints for the work and also provides me with the means to sell my assets if I wanted to. This is a standard which ensures that assets are on par with the standards that are expected of the engine, and run in an optimised manner.

Finally, forums like PolyCount and 3D media companies like Turbosquid give some indications of the kind of poly count that is typical of certain games.

The spreadsheet below details an artist at Turbosquid's experiences with real-time poly counts. While a lot of the assets on Turbosquid reach exponential amounts of detail, the 'real time' assets represent a significantly downscaled version of what is normally offered, while still maintaining a lot of detail.

For my work, as I have decided on a modular building approach, I need to consider how much the count is going to increase, just from modularity alone. From the spreadsheet, it looks like I should be aiming for around 100k per asset, which also depends on the type of asset, and its position in the world in terms of importance, and likelihood of being seen and scrutinised by the player.


Sources
Google Docs. (2017). Poly Budgets 2017. [online] Available at: https://docs.google.com/spreadsheets/d/1xJmSuBZbdMOlIcdXmO9hu46GJ7-wIiWCclJwimIOJa4/edit#gid=0 [Accessed 16 Dec. 2017].

Epic Games. (2017). Marketplace Submission Guidelines[online] Available at: http://help.epicgames.com/customer/en/portal/articles/2592186-marketplace-submission-guidelines-preparing-your-assets#Submission Requirements [Accessed 16 Nov. 2017].

Nelva, G. (2014). inFAMOUS: Second Son's Characters are 120,000 Polygons; 11 Million Rendered Regularly by the Engine. [online] DualShockers. Available at: https://www.dualshockers.com/infamous-second-sons-characters-are-120000-polygons-11-million-rendered-regularly-by-the-engine/ [Accessed 24 Nov. 2017].

Update on Concepts and Level Design

Originally I wanted to outsource any level design so that I didn't have to spend necessary time on planning and greyboxing a level.

While I have been able to find different concept art that would benefit my project, I was unable to find any premade levels. However, I was able to find a lot of pre-prepared level maps that I could use use and change to fit with my desired scope and still keep in mind my stretch goal.

Using an existing layout of a city block, I was able to block out the geometry of the level using Unreal Engine 4 and while it is an extremely large project, it allowed me to tweak specific sections and view them 'in-game' 

The biggest problem I face with using real-world maps and measurements is that, when viewed from either a third person person perspective, or first person perspective, the scale appears too small. This was a problem highlighted by Luke Ahearn in 3D Game Environments, where he makes reference to how racing games intentionally use massively wide roads to combat the limitation of racing in a two lane road.

In terms of concepts, I have gathered a variety of photo references, which I hope to reproduce using textures and modular construction and combining the two methods into a hybrid of modular assets and modular textures.

City layout taken from Akihabara. Overall, it is a lot to individually 3D Model, but may be possible as a stretch goal, using modular assets.