For absolutely no reason clear to me, render time dropped to 6 minutes (less than half!!) (EDIT: Now 5 minutes, wtf, is my computer attaining sentience right now or something?!) during a test today. With masks for animation, I got it as low as 20 seconds! I was doing some tests of focal length, because I think the scene looks like I filmed a miniature or something, the perspective seems… odd. Am I finally starting to obsess about quality and/or minutia, I wonder? Anyway, different focal length tests, for your appraisal:
35 (as before, lower POV):
Depends on what you’re trying to achieve here. Right now it’s just a random street with no focus (to my eyes, that is). All the focal lengths or POVs can probably work depending on what you as an artist are trying to show.
Also, when in doubt, you can use reference photos.
PS. Isn’t production already an iterative process as a standard: going from simple to polished?
PPS. At this level of detail, EEVEE could cut render times to “not-worth-mentioning”. Why not separate modeling and rendering process? Cycles has more believable lightning but since it’s a work in progress anyway…
Yeah, it probably needs full content use to make a good decision. Guess I’ll be going in that direction, then.
Re: Production, yes and no, there is a lot of iteration, but the steps kinda “trip over each other” unless carefully managed, which is probably why such a big part of a production is management in one form or another. I am trying to collapse those iterations down in a way, but I find it harder and harder to explain the concept. Basically, imagine everything done at once, but getting prettier. Yeah, that’s probably the worst way I can explain that…
As for renders, the render times mysteriously dropped A LOT. I think my cores were overheating last time, and slowed it down. Also, it probably depends on the machine power. But I look forward to seeing EEVEE in action when I upgrade!
Added traffic lights feel silly for not including those…
It is hereby confirmed that I have NO fashion sense, AT ALL. The guy needed clothes, though, and I thought I could inspire myself through Blade Runner 2049’s civilian fashions. I… I don’t know where it all went wrong…
Oh wow. I never even thought about how small the character would look from this camera angle. It IS an establishing shot, so it is how it is meant to be, but it is just weird seeing it for the first time…
Got 150 frames in the can and the limitations of version 5 are REALLY starting to show, especially problems with the character rig. But this is the learning experience built into my experiments, so I will wear it with pride as a batch of honor, dangit! Although it makes me uncomfortable to see his wrists bend like that… ouch…
The 10 second ‘Innocent victim’ shot for the version 5 trailer is now up. It feels weird to state as a positive, but a LOT of problems were uncovered in this process, not the least of which is, as many times before, the use of the NLA, which is still… sigh of frustration and sadness
But it’s in the can! Learning was done, pixels were rendered, pretzels were brutally consumed. Have a quick look!
Just put together the “alien morphing victim”, in two different tages of morphing, seen in the scenes with him in the bathroom and latr eating raw meat from the fridge (in the version 4 trailer). I still need to create the environments to make the scenes, but I think it looks kinda cool already:
On that note I will be making the rest of the contents in one big go, rather than scene by scene, as I had originally planned. The ‘Innocent victim’ shot was just to test out my new techniques. So expect it to be still shot for a bit before I animate the rest of the trailer as a whole… maybe…
Made the cyborg (who punches through a wall to grab an alien monster but gets ambushed). I have trouble differentiating faces, and I still need to get some kind of fashion thing going because they look a bit dull. But I got this far…
Updated the hallway and sewer a bit, although I think I’ll need some visible scoffmarks for the hallway, it looks too pristine… Otherwise, yeah, mostly just polished the version 4 files a bit (or whatever you do to a sewer…), but that’s the point, building the new by using what you already made.
Edit: Added the residential corridor, after filthying it up nice and yucky!
Edit: Added scoffing to hallway and reuploaded. Dirt on my screen and dirt on the image floor are pretty identical actually…
Had a couple of really crappy days, so productivity dropped. Instead, I finally updated the website, so a few of the things there actually work. Nothing new to really see, just got the videos in place and set the Tumblr up as a public development log. It’ll have what’s in here, but a lot less focus on technical stuff, and more of me rambling philosophistically about the meaning of it all.
Crappy days are continuing, but I managed to update the kitchen and plaza, although the plaza still looks a bit… dull. I lack the architecture / interior decoration skills to make it much better, so, for now…
The process I am trying to establish is to make fully functional projects in what you nicely call “simplified aesthetics”, which I call “quick and dirty”. Once they function in simple form, and ONLY then, do I start to push them towards either photorealistic or some other higher quality style. The holy grail I hope to create is a system for quickly realizing a project in simple form, and then turn those simple projects into more advanced ones.
I hope that made sense? People usually start looking the other way when I talk about this, so I rarely get to have a deep conversation about it
Edit: One reason for trying this approach is that I see sooo many who do amazing stuff and love explaining how to do it, but there is never really something they can use it for. I create something to use it for, and then hopefully learn how to make the kind of quality I see them make.
I tried a different design for the plaza. I think it has more room for improvement, but it definitely also needs that… Architectural (and fashion) design seems to be my great nemesis!
Made the new guard model. Not that big an improvement from the latter on the surface, but a few new things seem to work. I get the terrifying feeling that I am slowly hitting a ceiling of some kind, and will definitely need to go into learning mode after version 5, to become better at a ton of things. But at least I am learning about what I need to learn more about. Progress!!
