JinsariaKhavra
patreon
Overdue update on what's up
🕑 Added 2023-06-22 18:41:43 +0000 UTCTL;DR: Things are going well but there have been major changes behind the scenes. Content should pick up again soon with huge improvements in the medium to long term.
So, The very long story is that things have been a bit crazy behind the scenes but it's important now that I know where things are going that I let everyone know.
The good news is that in the long run this is going to be an improvement for content quality and the rate of content production. as well as improvements to the potential for the game development. And hopefully less stress on Aeva and I regarding the actual production of things.
Blender has been nothing but a headache the past year. fragmented versions where for every step forward and fixed feature there are two steps backwards and two more broken important components. In this case it's become incredibly difficult to see a path forward using blender in it's current state that doesn't include complete rebuilds of characters every few months like we have been dealing with basically the entire time of trying to make this work. Unfortunately fur seems to be constantly at the center of the frustration. A timeline of what we have been dealing with behind the scenes
The characters were originally built with 2.93 in mind then updated to 3.0 but blender has always had some amount of trouble with fur. In Cycles fur looks pretty good but the render times have been prohibitively long. Chapter 1 was proudly rendered in Cycles but a lot of the headaches behind the scenes were hidden with lots of careful post processing. Both OPTIX and OIDN de-noising in their out of the box state are basically out of the question, they both eat too much detail from the fur leaving it in a muddy,smoothed state and destroy the fine details. I purchased E-Cycles, K-Cycles and even experimented with Super Denoiser and finally settled on Turbo Render. Which got close to what I wanted to see but it created unacceptable artifacting.
The result with most options was that it rendered in a decent amount of time but left bright fireflies looking like radioactive fireflies or droplets caught in the fur or strange ghosting halos around fur or transparent elements making it look like the fur or even decals on things were superimposed poorly in photoshop. The only solution was hand cleaning up of every single image throughout Chapter 1. Basically every frame had to be hand cleaned of artifacts which took up to an hour of post processing by hand for each frame and most pages took an extra day of work just to clean up well enough to be presentable at the quality I wanted. This didn't improve with later versions and Cycles X only really made the problems worse, Adaptive sampling didn't help much and only giving the frame many thousands of samples truly cleaned up the problems. Render times were upwards of 22 minutes per frame just to clean them up in a reasonable way and that didn't assure that post processing wouldn't be necessary to fix occasional firefly clusters in the fur. I tried dozens of shader setups and optimizing lighting and the scene but the quality kept taking a hit that I wasn't comfortable with. We are a very small studio, Every asset is made by just Myself and Aeva by hand.
out of the box EEVEE wasn't offering much of a better option as it has terrible issues with how it handles the normals on fur, creating unsightly and unacceptable shadows at glancing angles, the only solution would be to massively increase the number of hair particles being rendered which is out of the question because even with 24gb of VRAM on my workstation GPU this was quickly flying outside of the limits of hardware. and due to the screen-space nature of EEVEE separating the characters into separate passes was out of the question.
In chapter 2 we initially moved over to EEVEE and Blender 3.3 using the excellent script by Ruaidri and Sentharn to encode custom normals into the fur but this had significant limitations and the implementation was buggy. Support for the script also quickly dried up with the last update being in august of 2022. The problems meant that animations were out of the question. once it was clear that support for the plugin would not be resuming I started working on a custom implementation using Geometry nodes, but initially there were problems with 3.3 that prevented UVs from being properly transferred to the fur meaning that textures would not align properly,this forced the transition to Blender 3.4 around page 7 and a complete rewrite of basically everything and reconstruction of the characters and fur systems.
3.4 has not been without it's issues either unfortunately. With simulate nodes and node physics being indefinitely delayed over and over there is simply no way to properly animate the fur in 3.4.
We dealt with the issues and purpose built rigs were created specifically to make things look better for some of the animations we did succeed in creating with 3.4.
3.5 was looking promising with it's release of proper hair nodes but making use of them proved to be around 100 hours of wasted time as they were simply too poor performing to have any potential of being used for full sized character grooms on a single never mind up to 4 characters on screen which was planned within just a few more pages.
To make matters worse EEVEE's performance has been continuously degrading. Shader compile times went from around 3 minutes in 3.1 to over 10 minutes for our scenes. shading tricks that used to work are starting to destabilize with no real solution offered other than waiting for the also indefinitely delayed EEVEE Next.
3.6 showed a bit of promise but many of the systems I used to build my fur rendering setup in 3.3 have been depreciated and a complete rebuild is necessary to make them work in 3.6 while the eevee degradation has continued, my screen space global illumination implementation has shifted from clever trick to hack and now returns flickering and strange artifacts in 3.6 and will also need to be rebuilt.
On top of this my health has been iffy, I'm still fighting with elusive cardiac issues (Prolonged QTc) and severe fatigue which often limits how much time I can work each week. And on top of that having very difficult to treat ADHD since my cardiac issues make my body incompatible with the majority of ADHD medication now.
having to completely start from scratch on Fur grooms,shaders,rigging, rendering setups while trying to match the old style to keep consistency, just to keep the comic moving forward has become too much and the solution was to simply stay on 3.4 and deal with the limitations but 3.4 wasn't issue free to begin with.
We finally hit a breaking point through May, deciding how to proceed with all of the limitations and frustrations, The decision was made that if we have to redo everything from the base geometry up at least one more time then we wanted to break the cycle of struggling with Blender's issues.
We invested most of our savings into a better solution. Buying a license to Cinema 4D, I have a lot of past experience with the software, I used to use it professionally for years in my past (pre-furry) 3D career, Unfortunately my experience with advertisement animations and architectural visualization only go so far with adapting the software to character animation so there has been a lot of learning plus the re-adaption time going back to the software after 5 years of mastering Blender as my daily driver software.
I'm already liking what I'm seeing and initial testing is going exceedingly well compared to the feel of trying to do this in Blender. Huge improvements to physics,rigging,fur,shading and lighting are all possible. I'm finally feeling excited to be doing this again. so far stability has been a tremendous improvement. I mean nothing is crash proof but blender's 20+ crashes per day were getting old compared to the 3 or 4 crashes per day of C4D. I determined that Redshift is probably not the best choice for what I want to achieve so along the way we chose Octane as the rendering solution due to it's speed,versatility and potential for expansion.
I don't think blender will ever fully leave my workflow as it's just too good at many things but combined with Cinema 4D's stability as a proper DCC I feel we are in a much better space.
Moving forward the plan is to finish out the remainder or Chapter 2 in blender but with the start of chapter 3 around page 20-25 we will be moving to C4D full time.
Just the potential for better softbody,cloth physics,and the exceptional fur middleware we have chosen opens way too many doors. and it is also offering us better animation capabilities and better integration with Unreal 5 so game dev should be faster in the long run too.
I'm looking forward to the reveal of what we have planned with this new potential in the near future. I really am sorry that content slowed down and dried up a bit but in the long term this is going to be the right decision.
Thank you for your time,patience and support.
