I’m sure you’ve asked yourself ‘how many polygons should this object be?’ at some point, especially if you’re involved in real-time asset workflows. To make matters worse, when you go to any online forum to ask for an answer, more often than not, the answer you’ll get is ‘it depends.’
This can be a tough question to answer. The issue is amplified when selling stock 3D assets that can go to all kinds of projects – you lose the project specific artistic and budgetary constraints that normally answer this question. Instead, you have to try and factor in every possible scenario where the model could be used.
At TurboSquid, we grapple with this same problem internally. After a lot of work, we’ve provided the framework for a solution in the StemCell initiative by saying that you should be modeling all the geo you need to in order to make the model as flexible as possible, but those are guidelines on modeling workflow and not hard numbers to rely on.
Recently, we reviewed our polycount cap for our guidelines on Real-Time content. At the moment, we list a hard limit of 100k polys, but since that was defined, racing games now regularly have models of cars that reach 10 times that, and characters on console games are reaching that number with greater regularity. To make matters worse, you might have that 100k character holding a tool that uses 5k polys, and anything beyond that is unnecessary for that object. No one number can describe a ‘cap’ today – the answer we ended up with was spreadsheet filled with dozens of model types, use cases, hardware generations, and special considerations.
It might not be completely accurate 100% of the time, but instead is a guideline – a very generous maximum cap on the item in question. Does it mean that your 100k pistol is game-ready? Probably not, unless that game is all about highly detailed pistols. The numbers here are provided with a philosophy that it’s easier to simplify and reduce the polycount of a model then it is to add new detail where none exists.
Please let us know what you think on the forums – we plan on can making it a widely available resource for artists, but we want this to be the most accurate set of figures as possible. Getting your thoughts on it will help a great deal!