NASA used an augmented reality dashboard for navigation of X-38. Astronaut, Scott Kelly, used this dashboard to report back to mission control from space.
You google it. Start to read about a hard surface, triangular polygons, retopology… Then you close your laptop. Now it’s even more complicated and you don’t want to run the project at all…
Don’t worry. We prepared a brief introduction so that you know the things you have to consider in order to be able to hire a 3D artist.
We will outline the main elements you should know or consider when looking to create your own models, environments or even set a team and establish a mass production process of 3D modeling for your next big thing.
Roughly speaking you can divide 3D models into two basic categories.
If you want to create some man-made object (scissors or Ferrari) it would require hard surface modeling. If you think about a living thing (octopus or Pikachu) here comes organic modeling. Main difference is that the forms and future movement of the object have to be natural (or superficial in case of a hard surface model), even if it’s a fantastic creature – our brains know certain flows a natural creature or object would have (grass also qualifies as an organic model, for example).
Usually, an artist specializes in one category. Our team, for example, tends to work with hard surface modeling due to the experience within the team.
While it’s not an easy task to create something new, it’s way more complicated to recreate a real object with photorealistic accuracy. You need some highly detailed photos and measurements (unfortunately, most of the times you don’t have a precise layout map) as a reference and plenty of time to do the job.
Artists tend to choose which of the sides is more appealing and interesting for them. Usually, working with organic objects requires more creativity and knowledge in anatomy. While hard surface requires strict precision and attention to detail.
Did you know that it takes about 10 days to create 1 photorealistic model of a household appliance (a vacuum cleaner)?
Our team has created the Augmented Reality application for ALLO (one of the biggest appliance retailers in Ukraine) in order to provide customers with access to 700 household appliances. Now that’s what we call hard surface modeling!
When you’re considering a new app using the 3D models in AR or VR there are a number of criterias to take into account. One of the most universal metrics to understand here is – the more polygons your models have the heavier it’s going to load up the app. Proper detalization can be reached using different technics. One of the most common approaches in “visualization 3D modeling” (usually applied in interior design visuals) is a large number of polygons.
We use the term “low poly” to describe a mesh that has a small number of polygons. “High poly” means (what a surprise!) a large number of polygons.
When you create a high poly model you get a photorealistic object that “weighs” a hell of a lot. Therefore you have a challenge – how to create a 3D model that looks like real yet weighs like unreal.
Going into specifics, there are a few approaches that help to do the magic. Our team usually uses a high poly to low poly workflow. This method allows creating a highly detailed object (4 million triangles is not a limit) and then converting it to low-poly (400 triangles e.g.) through retopology without losing the quality of the image.
Not only the level of poly is important to consider when you’re planning the future product. You also need to understand what would be the interaction with the object, which animations would be applied and whether there’s going to be a number of objects within the projection and a user would be able to move in between those – this could mean gradual loading of the models depending on how close to those the user is.
Mobile phones apply restrictions on Augmented Reality application polycount of 3D models. The capacity for the majority of devices is not enough for heavy models which should always be respected.
Facing the task of creation of 700 photorealistic 3D animated models for mobile application from scratch appeared to be the greatest challenge we’ve had so far.
Our solution was implying a very strict “model weight” limitations. We also used high poly to low poly workflow from time to time.
Internally we call our result “middle poly” models, as they are visually very detailed and precise yet maximum optimized.
Physically based rendering (PBR) is a beautiful method that serves to achieve the desired photorealism of the model.
PBR is a technic that takes into account real physical characteristics of the object and allows an artist to recreate those as realistically as possible.
Simply put, you have to combine a number of material properties (albedo, reflectance, roughness values etc.) to render it properly so it responds to surroundings properly.
As a business, you may require not only a set of models but actually an environment where you’d like to immerse your potential customers or present the whole scene as a part of your story. This could be a detailed manufacturing facility, a lab, showroom or even a city in either AR or VR.
Environments creation is a whole new level of 3D modeling. This actually requires additional roles for the full product to be properly framed and fully perceptible by a user.
While we could talk for a while about realism and the importance of thinking through all the details, thus letting your user dissolve in this new world; these are understandable metrics. What’s also important to cover would be a little bit deeper characteristic of a professionally made environment.
Environments should be easily adjustable and completely modular if you’re planning to work around the layout and variate it from time to time. A properly created environment can be easily managed by your team. Consider whether you need such an option beforehand, as it takes a completely different approach to its creation.
We created a simulation of the Zurich Airport environment in order to use this space to demonstrate our Virtual Reality environment creation capabilities.
Giving that we had to reproduce a smaller version of a real facility who would expect that the greatest challenge would appear to be not even that we had to understand the architecture and interior only from the photos we found online.
The biggest issue was to understand and realistically replicate the windows. Yes, windows. There was no reference of exact carcass available online. And our guys just couldn’t help the feeling that it wasn’t enough the way we did it in the first version. The frame was melting into the floor, which couldn’t be architecturally correct. Only when we found a 360 photo made by someone inside did the guys finally finish the model the way it was both logically correct and looking the exact same way.
If you need serious volumes of the models produced in strict timeframes, your processes should flow like clockwork. We found that such productions need very precise expectations for each model and team member. Such tasks require additional team roles, like artists team-leads and 3D art QA engineer, for example. Different stages of feedback provision for every model and specific flow of models refinement, final tests and release to production should be set.
This is a huge topic to be able to cover just in one chapter, but I felt it’s crucial to emphasize how important it is to have experienced management and professional feedback on models when you’re planning to involve more than 5 people into the 3D production.
Put very clear expectations to each member, draw clear roles, establish the full production pipeline and feedback loops within it, have testing stages, sync with the development team and agree on specific formats and test periods. As weird as it may sound, threat your 3D models as a usual part of the development process – it needs planning, demos, and full tests as well as retrospectives.
Try to project the exact use cases of your future product even before you start the 3D modeling process. There are so many things that would influence your artists’ future choices that you would want them to have all the information to make the right decisions. Starting from exact devices and platform that you would use, continuing with the number of objects in one simulation, possible full environment around the user and finishing with planning the animations and their complexity this object would have.
Plan beforehand and you will foresee some of the obvious mistakes in your product that uses 3D models and hopefully AR 🙂