A look into the future (Not mobile-friendly) (Patreon)
Content
Hey guys! First off, thank you so much for the overwhelming and still ongoing feedback on the recent animation! I'm quite happy to see the warm reception, especially considered how long it to animate and produce! (:
As previously mentioned, this might become a new series, as I'm already bouncing around ideas for the script of the next installment. For the near future though, I'll be working on some more bite-sized monthly animations, as well as some other things that I wanted to get into in this little journal. This is going to be quite a bit more technical than the usual posts, so here's hoping it won't sound like complete gibberish to most of you.
Short-term plans
The aforementioned sequel of Encounter will be the next big animation to come. This episode will be a lot shorter and go back to the more usual animation length of roughly 3 minutes as a set goal. Whether it will be more or less will show when I'm getting closer to completion, but it's supposed to be a project that can be pulled off in roughly 4 months. With this new animation I'm hoping to push my dated software to its limit, as by the time of production I'll finally have a new workstation that will be fully optimized for 3D animation. Luckily new GPUs are just around the corner, which means there was never a better time to upgrade! Throughout the next weeks I'm hoping to get my hands on a new RTX 3090, which should hopefully solve all the woes I've been having with video memory for the past year. VRAM is quite crucial, as it has a large influence on render time. Exceeding a card's memory makes rendering a frame significantly slower, and staying within those limits has become increasingly more complicated.
A large sacrifice that had to be made with Encounter was the skylight resolution of my custom-built skydome that basically consisted of over a hundred light probes directed at the scene. Each probe casts a dynamic shadow map, resulting in a large amount of shadow map data that simulates soft, diffuse lighting in real time without any raytracing required. The quality of the diffuse lighting is driven by the shadow map resolution, which had to be set to 1024x1024 per light, resulting in roughly 2GB of shadow map data. This was optimal for the 11GB of VRAM of the RTX 2080 Ti that was used to render this animation. I was originally hoping to render it at four times the resolution, but the resulting shadow data resulted in 8GB, which was too much for shadows alone, even if it would've given the environment a lot more definition, as this sequence demonstrates.
In the future this will likely become a non-issue, as 24GB of video memory should be more than enough for high-res lighting data, 8K textures, as well as Zoe's elaborate fur, which is going to get a significant upgrade as well!
Something that used to bother me with Zoe's previous fur shader was that it looked quite a bit flat and didn't quite have that full, clumpy look that you mostly see with animated characters of higher budged films. Throughout the recent weeks I've finally been able to take some time to redo certain areas that in my opinion needed a do-over. The areas in question were mostly above her chest, back, shoulders, as well as her cheek. This was finally achieved by making a second fur layer that uses much clumpier fur, somewhat akin to her tail.
I feel that these are adding a lot depth to the areas that make use of longer fur, especially to areas like her shoulders, which I always thought looked a little undefined and flat. In addition to the clumps, I've added another fur layer, which mainly adds fine singular hair poking through her tank top. These are mostly just visible from up close and quite subtle!
And now for possibly the best part! Zoe's new fur is now fully dynamic and reacts to her movement, just like the dynamic hair on her head and tail! This is admittedly quite expensive to simulate, but luckily I can turn all of that off during animation, so it doesn't slow down the performance within my program.
This new fur is quite a bit overkill to render for my current card, as it almost doubles the amount of individual fur. I feel that this will be a proper upgrade to her overall appearance though, and with the graphics cards coming out pretty soon, rendering these things shouldn't be much of a problem anymore.
Long-term goals
The future of computer graphics looks pretty bright! With RTX basically becoming more feasible this generation I've been starting to look more and more into raytracing. This won't certainly be something I'll be able to utilize anytime soon, but I've been experimenting with various renderers and even started porting Zoe over to Maya 2020, which was hell in high heels on its own. There's still a couple of problems with the rig, but I'm planning to eventually iron these out, as the new version of Maya offers support for a couple of amazing renderers that could bring future animations into legitimate Pixar territory.
One of those would be Redshift, which I've been extensively testing throughout the recent days. My goal here was to convert Zoe's basic materials to the native ones of this new render engine, which came with some decent results. The render times were still a little too long for my taste, but these were only done on a laptop with a GTX 1070. Redshift is RTX accelerated, so I'm sure that an RTX 3090 would yield significantly better results.
Apologies for the watermarks, I've only been testing this for two days, so I'm still running a trial version, hah!
Zoe is of course still missing her fur and her eyes also look a bit dull, but the improvement in material rendering should still be pretty evident in this image. Her fur is still by far my biggest concern, as fur has always been something that rendered ridiculously slow with raytracing. For this new version it would have to be remade from the ground up, so this conversion of Zoe is going to be a bit of a project that I'll keep running on the side. I doubt that I'll be doing any animating within the new software before 2021, but it's something that I slowly want to approach, as my current rasterized renderer will possibly be hitting its limit soon.
At last, a quick look at a raytraced Zoe. Don't mind the dull eyes, or her complete lack of fur. Oh well! :D