Progress Report - December 31, 2024

Hello, everyone!

It's the end of the year, and the end of my Blender explorations for now. And since it's the end of the year, I've decided to make this Progress Report visible for all Patrons.

The last thing I fiddled with was revisiting breast physics, namely collisions and jiggle. It's a poetic symmetry, ending the experiments where I began. This time around, I used cloth simulations, and they worked so much better than the soft-body I started with that it's frankly ridiculous.

Still not perfect, though. The biggest thing is I want to find a way to retain volume. Ideally I'd like to simulate the cloth bodies being filled with an incompressible fluid like water, such that any deformations would cause it to redistribute. In theory, that'd resolve the issues like the crinkling seen with the glass pane, and the warbling waves in the bounce test.

Blender cloth simulations have a "fluid simulation", but for some reason, whenever I turn it on in this test, the cloth just immediately explodes and scales up to infinity. I've tested it in isolation on a subdivided cube, just to see if it's a fundamental problem with the sim, and it seems to work fine. So there's something going on with this particular setup that the fluid simulation doesn't like.

I'm sure with enough tinkering I can get it, but for now, it's time to get back to work.

Tomorrow, I usher in the new year by returning to working on Overbreed!

My goal for this sprint is to finish off the Sombra sequence. I currently have 27 minutes of animation, and it's about 2/3rd the way through the storyboard. Assuming an even distribution, I'm expecting another 10 or so minutes of animation to be done before the sequence caps off around the 40-minute mark.

To be perfectly clear, Overbreed Episode 1 is more than just the Sombra sequence! Over on my blog, I made this handy little runtime map for the episode:

 What I am currently working on, and am hoping to finish the core blocking of, for this sprint is just Sequence C. It is the bulk majority of the episode to be sure, but it is not the whole video.

Indeed, if these estimates are all correct, Sequence C is only about half the full video. So when I do finally finish Sequence C, that will not mean the video itself is done. I just want to make that perfectly clear, since I know a lot of people think that Overbreed is only the Sombra sequence, and so they will see "the Sombra sequence is done" and read that as "the video is done."

With that being said, I am hoping that this sprint will be finished within a month, which is my ideal time split. Assuming that is the case, then once February rolls around, I will be shifting into making another short video, as per the workflow from 2024.

Hopefully this short video will actually be a short video, and only take 4 to 6 weeks to actually put together. Rather than the 15-minute 5-month production disaster that Pharmmetra was. Not to say the video itself is a disaster - I absolutely love how it came out. It's just the road to get there was... not good. I've written a post-mortem about it.

The February video is going to be one of two things. Ideally, the February video will be a Batgirl / Catwoman piece, written by everyone's favorite voice actress InsideIncognita. A woman of many wonderful skills, not only is she a fantastic voice actress (and a beautiful singer to boot - one day I will put her singing ability to good use!), but she is also an avid writer. We've been talking about collaborating on a piece ever since Blue Star Episode 3's production, so for over half a decade at this point.

The collaboration I did with DrJavi for the Claire/Jill video Refuge came out fantastically, and was a supremely fun experience for me. It feels nice handing over the creative reins, and I just love to see how other creative minds spin. Innie has been wanting a Batgirl / Catwoman video for just about as long as we've been discussing collaborating, and so I feel it only makes sense for the stars to align and for that to finally happen.

She's been busy hacking away at her manuscript for the project. I told her to not concern herself with length or feasibility, and to just write what her heart desires. And boy howdy, has she been writing.

As January draws to a close, assuming Innie's manuscript is complete, we will begin the process of ̶b̶u̶t̶c̶h̶e̶r̶i̶n̶g̶ ̶i̶t̶ editing it into a malleable form. The major steps there will be converting it from its current novella form into a screenplay, and then cutting it down to size. I'm aiming for 5 minutes, 8 minutes at the absolute maximum. I have no idea how big or elaborate Innie's manuscript will end up being, so I have no idea how difficult of a process that will be.

But we'll cross that bridge as we get to it.

In the event that Innie's manuscript is NOT ready in time, I do have some contingency projects in mind. Right now, I'm eyeballing a Mass Effect Jack x Miranda project. I know some people have made full videos between the two, but I don't think I've seen anyone take the particular direction I have in mind and build anything more than a loose loop out of it.

Looking ahead, after the next short video is done--whether it be Innie's Batgirl / Catwoman project, or a contingency project--I intend to return to spending another month exploring Blender.

My end goal is to eventually move over to Blender to fully take advantage of the various simulations it has available. As I expected, the technical side of actually wiring up all of the simulations is pretty straightforward - you can get a tiddy cloth sim up and running inside of 15 minutes, with very limited Blender knowledge.

The problem is in the details. The simulations have a lot of settings, each of them being extremely sensitive and having profound effects on the final result. Minor adjustments in any one of them can completely alter the simulation.

I knew this was going to be the lion's share of the time utilization, but I didn't anticipate just how much time it takes. And that's really what it comes down to: time.

Tuning the simulations isn't difficult. Anyone can click a text box and type in numbers. It's just a matter of mind-numbingly boring trial and error.

Once I do finally have the settings dialed in though, I fully intend to write some Blender Python code that will hard-code and automate them going forward. The jiggle physics I use in SFM were the same process - a few hours of trial-and-error adjusting numbers until they behaved just right, and then writing a 3dsMax script that automatically generates the code for all future models to use.

For that month of Blender experimenting, I intend to split it more or less down the middle. I want to spend the first 2 weeks of that month exploring further systems, to get things viable. Right now, the big things I want to explore are things like ropes and chains; writing an SFM equivalent Blender script to automate setting up IK rigs in Blender; and explore writing a better animation setup that more closely emulates SFM's Animset Editor for Blender, with proper arbitrary grouping, and an ability to quickly swap between animating bones and flexes.

Then the back two weeks will just be me hitting my head against the wall as I try to dial in the various settings for simulations. I'm pretty happy with the physics-driven hair that I showed off last week, so I think I will start with with that and use it as a basis for experimenting with writing Blender Python code to automate it. Use Anna's braids as a basis, hardcode the settings into Blender Python, and then apply them to Elsa's ponytail and see how it works.

Once that is done (hopefully won't take long), I move onto the less fun stuff. I'm going to work my way from least-work to most-work, and so I want to start with dialing in fluid sim physics. I want to see if there are any third-party plugins for Blender to serve as an alternative for Blender's high-viscosity fluid simulator, because I really do want high-viscosity, but fuck me if that sim doesn't take forever. If I can find a faster third-party solver that focuses on high viscosity, that'd be great. If not, then it is what it is I guess.

After the fluid-sim stuff, I want to return to non-physics-driven hair. This is an experiment I never showed off publicly because it did not go well at all. I think I want to use the same base that I did for SFM, and use springs for the base motion. I looked around, and someone recreated SFM's jigglebone system in Blender (the addon is called Wigglebones). And that's a great start, but it has the same limitations as SFM's jigglebone system - limited reaction to gravity (you ever flip a SFM character upside-down and see how the hair just kinda gives up halfway through falling down?), and importantly for me, no collision hulls - the hair just clips right into characters' faces and bodies, which I very pointedly want to avoid.

The ideal solution to that is a spring model with a collision hull. One that, under initial conditions, does not deform the hair in any capacity - but as it deviates from initial conditions, EG the character moves around, reacts accordingly to gravity and uses the collision hulls to avoid clipping into the character's face and body.

I don't know if Blender has the capacity for that. That's what I intend to find out.

Honestly, I don't see me returning to tuning the jiggle physics in this next Blender sprint. I think I have enough on my plate with all of this.

And then after that Blender sprint, we repeat the cycle again: Overbreed, short video in SFM, Blender exploration.

I'm not going to project into the future any further than that, at this point.

So yeah. That's where we are, and that's where we're going. Next week's Progress Report will be the first of 2025, and hopefully we usher in the new year with some positive progress on Overbreed!

Until next week, everyone!