bloom - submission post










Reflective statement

Over the course of this project I forced myself to learn and use as many programs as I needed to. With many games companies using many different programs I find appropriate to learn as many as you can because you don't know what they will use. In this way I feel as if I'm pushed for time to take in a lot of knowledge rather quickly. Having said that, I am happy with what I added to my repertoire. Orantrix, dDo, ART, and Marmoset Toolbag 3. 
I felt the most uneasy towards the beginning of the project in which I couldn't decide exactly what type of design to go towards. On top of this I realised I was finding it hard to distinguish between good design and personal aesthetic preference. Though it's okay to have them crossover; I thought I was relying too much on what I tend to gravitate towards. However in the end I have become rather fond of Sawa's design. Whether or not that design was properly fit for 3-D is another thing. I found when doing the retopology that the design of the suit was not exactly efficient as the subtle silhouette difference made for a lot more work with not a lot of visual "pull". I was also completely unsure if the retopology I had done would animate properly, no matter how much research I had done. But this was more an issue of my inexperience than obliviousness. Speaking of which, I had conducted a great amount of research both beforehand and at the time in order to reduce the amount of grandeur mistakes made. I did however make a lot of mistakes down to simply not knowing a small part of the pipeline that would nevertheless stop me from moving forward. They felt like good mistakes, though. Ones that I knew I would not make again and thus help me speed up the process in the future. For example I had a problem with Marmoset Toolbag in which I didn't realise it did not accept material IDs. Whilst it seems insignificant, it actually became rather difficult trying to figure out how to break the mesh up in order to have the skin and body elements separate to texture them properly. 

As an over-arching theme; I don't believe I managed to optimise my character that I made in 3-D well enough. But this would've been to do with becoming overwhelmed with the workflow and thus slipping up on the cleanest way to do things. Although I must say, and this is partially because the character was for games; at no part did I feel as if I was making decisions that could be classed as "cheating." In that regard; I didn't cut corners because I couldn't figure something out, but I also didn't cut corners in a good way - for optimisation. This will be something I'd like to explore in the future.

As a last note, and perhaps I'm shooting myself in the foot here, my one major gripe with the presentation of my project is the pose I gave Sawa. Though I used reference from a real ice skater I don't believe I chose the right one. Aside from that, this project has been an eye-opener for what I can achieve by myself as often I tend to subdivide what one person's role is, akin to how the industry works. This project has given me more confidence with what I can achieve by myself and also a good idea of where I stand skill wise and what I need to improve on.

bloom - skin and hair~




so i had no problems importing from dDo into marmoset toolbag 3. I worked on the skin shader a great deal today, as well as the hair and eyes, though I'd like to continue tweaking the hair. From these screenshots I see I need to move the lower eyelashes down a little too. She's going to look a little wide eyed until I get the face rigged up so I can lower her eyelids. 

Actually I had a problem where the gloves and the back of her are not separate pieces, unlike the head. I did not realise Marmoset Toolbag doesn't support Mat ID's, thus I had the choice of either separating the back (easy) and the skin part of the hands from the gloves (annoying star shape). I realised instead I can manually take all the maps I made in dDo and create additional maps for the skin by hand, masking off the SSS, emissive layer, and detail normal maps from the cloth materials. There is a little less customization but it still works. So now I know for the future. : )



bloom - texturing part 2


thought i would upload another preview now that i have the bulk of the objects textured. this is previewed in dDo, the next challenge is going to matching it properly to marmoset viewer + figuring out the skin and hair. when that is sorted out i'll go back to the holsters and guns.

Bloom - dDO texturing



i thought i'd just share some of the progress i've worked on today with the texturing in dDo. Using material IDs, normal maps and AO. cavity maps were generated from the AO maps thanks to dDo being clever! i think i'd like to tone down the latex on the long glove.

Bloom - Normals & AO bakes




Finished baking normal maps and AO maps. Used xnormal. Tried handplane but with worse results. The gun was annoying because I couldn't explode the low poly (due to the high poly having details that I did in zbrush ontop of the original maya hp version) so I had to bake each piece separately so that the ray casts didn't interfere with eachother. So I just took the parts and combined them together in Photoshop for both the normals and AOs.


also, i sculpted the iris detail. actually i did this on my les parapluies de cherbourg image from year 2 but from a drawn texture. 

Bloom - Retopology


normal map + uv preview
I finished the retopology and the UV's. I'm just currently moving shells into the 1-1 space rather than using separate UV sets as I'm not entirely confident with what programs support them. i decided i'll be using dDo for texturing as it's basically a built in bridge with Photoshop and works seamlessly. I'm baking normals and AO maps right now. Going to bake & work at 4k then scale down to 2k in final.

The model is overall 80k tris. (including holsters + guns) I worked to FFXV's polygon count of around 100k polys per character + 20k allocated to the hair alone.
Hair makes up about 19k on mine.

The two suit parts, armbands, glove, and bow part are one combined mesh, and combining the fancy raised part with the vertical striped underlayer was probably the most challenging thing to do and was laborious to say the least. As I worked on it I began to think that it's probably conclusive of not much thought on my part going into the design process. If this was a lower poly character I would not have been able to get this same raised effect in the silhouette. (One that is minimal at best regardless, and especially from far away) Learning experience ! Furthermore I was worried about baking a normal map, but I tried especially hard to respect smoothing groups. The edges of the model where the objects usually break up have hard edges and are separate UV islands. Thus I am having no problems with baking. (when using cages properly, that is)






example, low poly left baked normals on right (not final bake version)