Skip to main content

Asset Optimization

A

bout a year ago I posted my top tips for UV unwrapping (see here). Since then, I’ve FINALLY started working through my backlog of models and texturing them (some of which I created five years ago…where has the time gone?!). While prepping these assets, I spotted some un‑optimised geometry that needed fixing. This then felt like the perfect opportunity to share a few more tips on how to shave off the poly count.

Like UV unwrapping, model optimization is a necessary evil. However, unlike UV unwrapping, you can adopt what I like to call the 'polygon janitor' mindset whilst you build the model.

Here are five tips I've found particularly helpful while mopping as you model:

  • Pre-visualization: Before you start moving pixels, picture how your asset may look like untextured in wireframe view. Visualize the curves and edges. Is this a hero prop or a background asset? Are there specific areas that need more detail? How close will the viewer/camera get? If the answer is 'far away' or 'low‑impact', you can afford to keep the geometry pretty simple.

  • Poly-Density and Hotspots: As you add hundreds of points and vertices, look for dense areas and hotspots like on edges and curves. Constantly ask, do I really need a crisp smooth bevel or can I get away with the bare minimum? For a videogame, optimization is absolutely crucial for performance. Whether the model is for VR, film or an ad, a clean mesh is a happy mesh!
  • Interior Faces and Hidden Verts: Ask yourself, will anyone ever see the interior faces of this object? If not, delete it. Interior faces actually make up the bulk of unnecessary tris. Targeting these, you’ll cut the count without visual loss. Also, look out for overlapping vertices. In Maya, utilize the 'Merge Vertices' tool with a shallow threshold. Be thankful such an automatic wonder exists; saving you a lifetime of manually merging two verts one at a time with the Target Weld tool.
  • Over-attached Points and N-gons: Trim the excess geometry now. Future you will be very VERY grateful. Try to avoid over-attached vertices. Keep the limit of six edges for one vertex for best practise. Whenever possible, replace N‑gons with quads. Quads split cleanly into triangles on export, and keep the count reasonably predictable.

  • Purge and Purge Aagin: After you think you have finished reworking the model, run through the steps again. Even better, go slightly beyond that. You'll be surprised how on each pass you can uncover a better layout.  Optimization is an iterative process.

    Underneath is a before and after comparison of a recent revision. These sorting table wheels were at a count of almost 2,000 tris. After following my cleanup, it was down to 1056 tris. Don't be afraid to keep refining. Remember the saying "Good, better, best. Never let it rest. 'Til your good is better and your better is best."!

     


    Jason Bartlett ©



    Next post, Sorting Table renders!


    Thanks for reading :)

    Comments

    Popular posts from this blog

    Crab Pot

    For the next step of the modelling process of the crab pot, I was challenged with another huge hurdle ( to say the least). This stage was where all the painstaking problems crawled in again.  The rope. I thought making some bendy cords would be simple enough, but believe me, it was anything but. There aren't many tutorials, covering the whole modelling process to the end 'game ready' product, for rope. So, like previous obstacles I faced in the modelling stage, I used multiple methods and trial and error to find the best path.  Actually, creating a tube following a manually placed path, was the easiest part for me. I had picked this step up while at University. However, I had to manually place and move each Bezier path to form the guide lines, for yards of rope, in a natural and realistic way. Hundreds of them. I spent more hours than I would have liked, setting up the paths to guide the rope along. I kept asking myself that there must be an easier way to tie...

    Render Tests

    An important element in the conceptual stage is to not look at a problem from one perspective. It's always best to get multiple views to achieve a greater outcome. To reach an end goal is one thing. The journey to that end goal is something else entirely. One of my favourite quotes by Carl Friedrich Gauss states "It is not knowledge, but the act of learning, not possession but the act of getting there, which grants the greatest enjoyment". I totally agree with this and it applies to game development. Creating something big or small brings its challenges and the act to reach the finish line is where the magic happens. I hope to experiment in several styles, as would any other artist working early on in game development; focusing on diversity and demonstrating imagination. Jason Bartlett © I created the above image, of a  bow mast, using the toon shader in the render settings of Arnold for Maya. I really enjoyed creating the style, was simple to se...

    Timeless Terror: Jaws @ IMAX (50th Anniversary Re-Issue) Review/Analysis

    T he 50 th Anniversary Re-Issue IMAX screening of Steven Spielberg's  Jaws  (1975) was not just a revisit to a classic, but a thoughtful re-education and celebration of cinematic history. From the moment the MCA Universal logo and the iconic JAWS title card graced the colossal IMAX screen, the experience justified every penny. The stunning clarity plunged the audience into Amity Island’s perilous summer like never before. The enhanced resolution revealed details previously lost to small screens and the compression/conversions to old VHS and DVDs. A prime example of this is immediately apparent in the opening sequence. The day-for-night shots where Tom is running after Chrissie on the beach, is remarkably clearer. This allows for a much richer appreciation throughout. As someone with deep admiration for analogue artifacts, particularly works shot on ARRIFLEX or Panavision cameras with anamorphic lenses, this IMAX presentation elevate...