(???) 1 Gold Portal - Last WIP

Phew. This is the last WIP for this animation, as you can see it's pretty close to the final look, but there are still things to do.

Last project with the new model

First, I want to talk about the new model. This project is the last one in which I have used a new model. It's not because it's a failure or bad, nope. Considering how raw it is, it has done a really good job in CA10, this animation, and the previous short animation. The reason why its last is that I began getting critical issues, and it's quite hard to fix them. Shapekeys are a great thing to fix most things and I was underestimating how useful they can be, especially when working with external software. But they also produce insanely inaccurate results when we're talking about model collisions.

So... It's time to return to continue working on my Hoodwink model. I've gathered enough info and issues from the last projects. There is a TON of things to fix and redo, but after I fix them, we'll finally will have a good model to work with. I guess there will be another series of test animations after I finish working on most critical issues, but we'll see.

Unreal, and this animation

Uhh… I didn’t thought of how hard it will be to make actual animation instead of test one in Unreal. In previous animation I was trying to get at least somewhere considering how bad things went. And I didn’t really tried to fix issues and push it to Blender-quality projects.

In this animation I’ve tried to use most things I can easily do in Blender. And while it didn’t went horribly wrong, it was a hard challenge for sure. Main problem is to export whole scene with physics and simulations into Unreal so everything work correctly. Remember when I said I’ve found a way to export my scenes into Unreal without losing details? It did work but with some quirks. It does export everything correctly as long as it’s something simple, on Source Film Maker level (I do not say that SFM is bad, but it’s limitations make most projects look basic and easy to export anywhere). For my case, I regularly modify meshes and use in-Blender tools to make specific stuff and they don’t really like to be exported anywhere.

Now to quirks - There is a lot of them, but I’ll focus on one. Exporting Hoodwink, her body more specifically. I made her body low-poly on purpose as animating subdivided models is a pain in ass that requires powerful CPU. This leads to sharp corners where they shouldn’t be, like on her ass. When I exported her into Unreal I was faced with issues related to lack of polygons. The solution is simple - just subdivide model! Only if I was working only in Blender… Because I need to export her into another software, I cannot use this simple solution. Okay, then just subdivide model and export it separately (Model mesh and animations are stored separately, you can change mesh in any way you want as long as the bones in armature don’t change). Good, this one does work… One thing - all modifiers erase shape keys when applied on export. This model is HEAVILY rely on shape keys. So I got into position where I need a subdivided model as low-poly looks horribly, and at the same time I can’t easily do it because I need shape keys. Gladly I’ve found a way to do it, but it was a long journey.

I want to mention high-caliber solution to every troubles I have here - Alembic files. First, note to 3D Artists out here who use Alembic’s to export their works into Unreal, Houdini, etc - I don’t want to use it in every scenario, as Alembic files made by Blender are insanely large and tend to have issues with skipping frames, wrong vertex positions and etc. In my experience they are unreliable and they are pain in ass when you work on animations with 300+ frames. Now to simpler things. In this animation I also used Alembic to export her cape and Vulperas stick. Due to how they rely on modifiers, exporting them like everything else wasn’t an option. This led into issues inside Unreal. There is two ways to import Alembic into Unreal - as Geometry Cache and as Skeletal Animation. First one always work correctly but it doesn’t work with Motion Blur which is requirement for cinematic. The second one works only if exported mesh is consistent and doesn’t change its topology throughout playback while supporting Motion Blur because of how it works. In simple words, it takes base from one frame and generates shape keys for each frame. In results you have a mesh which deformations were baked in shape keys for each frame. Issue with it is that it’s generates a ton of “useless” data which can take a lot of space. The solution for that is to reduce amount of shape-keys generated for a price of accuracy.

With how hard it is to move scene into Unreal, I’m considering making a “Guide” for those unfortunate souls who decided it will be a great idea to render their works in Unreal instead of Blender.