So I’ve been having a hard time meeting the poly budget for this project in the quest to retain as much of the silhouette as possible. The poly budget of 7k makes me struggle quite a bit cause what I ended up in my retopo is rigid and creates awful baking issues. Please check out the images of the low and high-polt below.
Essentially, I need some advice/help in terms of how to correct my topology so it will capture as much as possible from the high-poly. When I tested things, the most problematic objects were the face, as well as the body.
Note: I’m not asking help for a paid project. I’m learning the game pipeline.
Hi, i think if you delete any surfaces that are hidden/not showing underneath/inside the clothing and props that will reduce the count dramatically, also get rid of the triangle on face, shoulders and hands.
The entire project barely has hidden geometry except for the lower part of the shirt(?) since it connects with the upper part, including the belt, armor as well as the arms in order to have a watertight mesh.
I think you could reduce polygons in non-deforming areas to free up budget- for example, the sword doesn’t need the middle edge loop, it’s a hard object. The chest probably isn’t going to deform much, so you could probably half the number of quads there. Your bag has more polygons than your character, it looks like. Why is it so dense around the edges? You can reduce that significantly
The middle edge of the sword is there because it’s meant to add a ‘triangular’ look to the blade. Otherwise, it gets flat. I was also thinking for reducing the quads of the chest there but then when baking I have huge difference between the high and low-poly. Other than that, I’ll see what I can do with the bag.
Any suggestions about what I can do with the face, though?
Yeah, you have a point, thanks! I’m trying to reduce the polycount as much as possible, however what bums me out in terms of baking are the face and the body, especially these parts:
I imagine sleeves must be difficult, but what’s wrong with baking ears? They look so flat, almost seems like you could get away with no geometric details at all there.
For the inside of the ears, you could likely remove all the polygons and have the detail be entirely in the texture. When baking a sculpt, concave areas that are entirely surrounded (holes) are more forgiving because they don’t affect the silhouette.
Now I get you. Thanks for going through the trouble to do a demo for me. What I aim though, in the best case scenario of course, is to retain the depth of the year or have it 3-dimensional instead of hand-painted.
Eventually, you’ll have to make a choice between “depth” and “7K polygons”. Realistically speaking, 7K polygons is not enough for depth. You’re lucky you have the option to do hand-painting at all, rather than being limited to per-vertex color. You can get a surprisingly high amount of detail without painted textures, just with good normals, and with both, you can do basically anything. Pretty incredible characters have been made with way less than 7k polygons:
With the polycount you are aiming for, some sacrifices will have to be made somewhere on the model. If you want depth in the ear, you could have the inside of the ear converge to a single vertex instead and you would still have some impression of depth (especially if this is a case where you are baking from the high res model).
I hear you, guys. I will give my best to make it happen. Meanwhile, I’d like to ask you how do you find the facial topology? For reference purposes, the mouth will be closed (not animated)?
I think the face has a good topology, though a bit heavy at places for what you are going for (it’s probably responsible for a large part of your polycount). Low poly characters have somewhat more permissive rules for topology, because they won’t be subdivided. Maybe some of the small polygons around the border of the hair could be merged, as well as some parts of edge loops that won’t affect the silhouette too much.
What I mean with merging part of an edge loop:
When reducing a model, you might have thought about selecting and dissolving entire edge loops, but there is nothing that forbids you from merging only less important sections and keeping the more important ones:
Ex. in your case, the forehead has lots of vertical loops, which could probably be merged by groups of 2 or 3 without affecting the silhouette much. Except you only merge them until they reach close to the eyes and then you allow them to expand so the face can keep its detail. Triangles don’t really matter, as we won’t be subdividing the mesh.
The part i’m questionning is more the hands than the face. If you have made the fingers cubic, why is there a middle edge splitting each finger that makes no difference for the shape?
How close will you be getting to your character in the game? Sometimes, it’s easy to get wrapped up in details that you won’t be able to see when the game is being played ( and in cases where you have cinematics or close ups, you can use a “hero” model). Some things to consider: Will you see inside the mouth? Will the fingers operate independently (or will a “mitten” hand work?). Someone else mentioned it, but if an edge doesn’t contribute to the silhouette, you should consider getting rid of it (unless it’s needed for deformation). Also, consider your animations: if they are fast and explosive, you can usually get away with somewhat cruder geometry - because the deformations will be covered with motion blur - or just speed. Also consider a “segmented” design if it makes sense - i.e. don’t knit the shoulders into the body, use “hot dog ends” to make the deformations less critical and hide this kind of joint with your textures.
etn249 brought up some going points: Details inside a depression can be achieved with a normal map very convincingly. Small edges - like along the shoulder armor and wrist cuffs could also be achieved with a normal map. On the head/face, you have a lot of fine-edges that could be represented with normal maps - unless you are using those edges to delineate material borders. Fine bevels are usually not necessary - look into “Korean Corners” for solutions. On the hands, you have a top/bottom split line that could probably be deleted. I see a lot of small detail edges that are useful when rendering with Cycles, but in a game, not necessary - use maps.
Thanks for the specific points you address and insight you shared, @JoeW! I’ll consider it carefully. Meanwhile, it was very foolish of me to not post back view of the game res version of the character. I’d humbly ask you to share your thoughts if you see problem areas, especially the back side of the ear.
PS: I haven’t implemented the above mentioned suggestions yet, so this is still the last state I left the mesh before posting this thread.
One thing to keep in mind: ALL meshes will be triangulated by the game engine via a process called “tri-stripping” - so trying to maintain quads is pointless and can actually be a detriment to how your model “reads”. A quad can be “folded” due to the non-planar nature of organic models - so - that quad can read as a ‘depression’ when you want it to read as an edge. Game engines will almost always make bad decisions when it comes to triangulating your quads, so, you will want to triangulate your own to maintain control of the look of the model. (NOTE: Even Blender triangulates quads internally - you just can’t see the “hidden” edge that splits a quad into two triangles - but that edge can be “flipped” if necessary).
In this image, you have two quads at the bottom that have points lifted to twist them out of plane. Notice how in one case, the “crease” shows up across the top, and the other, across the bottom. This shows the “hidden” edge. The top two quads show that manually triangulating the quads can make a huge difference in the shape (silhouette). A good habit for lower-poly characters is to rotate the character slowly and look at the silhouette - and where you see divots and folds that you don’t want, you manually triangulate those areas (it’s not a bad idea to triangulate the whole model - once your happy with the topology).
In regards to your existing mesh, I’ve taken the liberty to make some suggestions. I’ve indicated some areas you will want to manually triangulate (lines) some edges you could probably delete (remember: overall polygon density should be consistent unless it’s critical for some reason. Eliminating those “corners” on the shoulder armor, for example, won’t really hurt the shape relative to the rest of the model). The polygons with “dots” are ones I would consider getting rid of. The “?” indicates something that I don’t know what’s going on - it looks like a massive n-gon … maybe you had that edge loop selected and hit “F” by accident? You may want to go into Edit mode and select n-gons via the Selection menu (Select>Select all by Trait> Select by sides> Greater than)
As a side note, I see a weird vertex in the back of the heel - like there is an edge that’s subdivided … might want to clean that up as it will save triangles Also, the hands are pretty dense - will you be able to see them that well in the game? (i.e. is the density justified?)
Also, the hands are pretty dense - will you be able to see them that well in the game? (i.e. is the density justified?)