Static camera scene: High poly VS Pre-rendered
null
Hi everyone!
First of all, I apologize if I have put this thread into incorrect section. Second, nice to meet you!
I am just an amateur 3D hobbyist, started using Blender maybe few months ago. Anyway, I have a question for much more experienced modelers than me
I am planning to make a static camera survival horror (Resident Evil old school style) over the next 5 years (I am a dentist, don't have much free time...). I have found this site and similar ones to be quite useful for free source models to create "static" props in a scene: https://archibaseplanet.com/gdl
These are CAD models and are mostly high poly. For example, a table lamp may be 30000 triangles, a full bed may be 150000 triangles.
So far, the work flow to convert CAD into "game" model in Blender was as follows:
1) Scale and rotate appropriately the object/s
2) Remove doubles, convert to quads, delete un-seen faces
3) Place seams via edge loops and UV Unwrap
4) Apply appropriate materials / textures to each part of the model
5) Add the object to a scene, add other objects (all objects are high poly)...
This takes time of course, but not too much. Time is very important factor to me.
If I would just save and export this scene with high poly objects, after combining all objects in one and baking color texture only, I would have a single mesh with about 500000 - 1 million tris. This would be a single scene + character + weapon + enemies (1 - 4 at a time). So total polycount per scene would be more than that.
Baking everything in a scene would take ages - re-topo + baking = way too much time and effort.
Considering that PS4 could output 30 million tris per scene in 2014 for a certain game, should I just say that polycount does not really matter anymore and get on with it? Or do I have to make low poly + bake from high poly? What is the advantage of that if I use much much more time and GPUs these days can eat millions of polys easily? Obviously, for character / enemies / weapon / items I would do low poly + bake.
If this plan is flowed, would I just be better of rendering a scene to picture, then use layers or depth buffer in a game engine to fake 3D? Then have 3D characters / items on 2D background, the 2.5D approach?
Thanks for your patience!
First of all, I apologize if I have put this thread into incorrect section. Second, nice to meet you!
I am just an amateur 3D hobbyist, started using Blender maybe few months ago. Anyway, I have a question for much more experienced modelers than me
I am planning to make a static camera survival horror (Resident Evil old school style) over the next 5 years (I am a dentist, don't have much free time...). I have found this site and similar ones to be quite useful for free source models to create "static" props in a scene: https://archibaseplanet.com/gdl
These are CAD models and are mostly high poly. For example, a table lamp may be 30000 triangles, a full bed may be 150000 triangles.
So far, the work flow to convert CAD into "game" model in Blender was as follows:
1) Scale and rotate appropriately the object/s
2) Remove doubles, convert to quads, delete un-seen faces
3) Place seams via edge loops and UV Unwrap
4) Apply appropriate materials / textures to each part of the model
5) Add the object to a scene, add other objects (all objects are high poly)...
This takes time of course, but not too much. Time is very important factor to me.
If I would just save and export this scene with high poly objects, after combining all objects in one and baking color texture only, I would have a single mesh with about 500000 - 1 million tris. This would be a single scene + character + weapon + enemies (1 - 4 at a time). So total polycount per scene would be more than that.
Baking everything in a scene would take ages - re-topo + baking = way too much time and effort.
Considering that PS4 could output 30 million tris per scene in 2014 for a certain game, should I just say that polycount does not really matter anymore and get on with it? Or do I have to make low poly + bake from high poly? What is the advantage of that if I use much much more time and GPUs these days can eat millions of polys easily? Obviously, for character / enemies / weapon / items I would do low poly + bake.
If this plan is flowed, would I just be better of rendering a scene to picture, then use layers or depth buffer in a game engine to fake 3D? Then have 3D characters / items on 2D background, the 2.5D approach?
Thanks for your patience!
Replies
Gives me a piece of mind and now I can focus on 2.5D scene
I can then add details say in GIMP, like blood stains?
So indoor scene creation:
1) Model the room, unwrap, texture
2) Import furniture etc, adjust scale, optimise topology (quickly)
3) Place seams, unwrap if using textures
4) Create PBR materials or install them (plus special materials like rusted metal, fogged glass, stained wood, etc )
5) Set up light sources and cameras
6) Render to picture, also render depth map (plus any animations of a scene)
7) Add things like stains, blood in 2d drawing program??
Also you dont have to do any unwrap (depending probably on software you use but most packages (3ds max, etc.) can do easily without unwrapping.
Im by the way also thinking about making a game (adventure something like point and click more or less) and was also split between full 3D and 2.5D (3D characters, 2D pre rendered scenes) and also decided to go with pre rendered. It just much more easier/faster not matter what some "experts" on gamedev forums tell you :-).
Also i will add aditional question which will probably affect you aswell. The downside of using pre rendered backgrounds (if you are going for more realistic look) is scene animations (door opening, machine working, tree moving branches etc. in wind). These have to be rendered frame by frame. And if it animation "sprite" is big say 500*1000px image) and has a lot of frames, it could AFAIK more HW heavy/demanding than even the latest 3D shooting blockbuster :-).... But without any pre rendered animation the scenes/backgrounds/room look static.
So i wanted to ask people around here with more experience:
Is the current mainstream HW of PCs can handle for example 72 frames (500px * 1000px) animation? (tree waving in the wind)? Or is it still too much for current mainstream PCs?
Ye, you are definitely right about large animated frames. I have found this post about Resident Evil Remake (2.5D) quite useful:
https://www.gamefaqs.com/boards/535836-resident-evil/53920962
Basically they use 1 - 2 frames for most scenes (including flashing lights from thunder), so I think we can get away with a couple of full size frames
I will read the article... Does anybody have else have knowledge about this issue?
Alastor¨: I didnt quite get what do you mean that you need convincing color shine/reflections etc. You can use of course different types of maps (diffuse, refl, bump, displace etc. even on NON-unrwrapped mesh (when you are going for prerendered mesh)....
Wait... I don't need to unwrap??
Comping layers/paintover/photobashing. Anything goes. Thinking that you have use the 'pbr' workflow for this is limiting yourself. You don't need a normal map or an AO map. You can use geo and an AO pass. You can render out separate spec or reflection passes and use paintover. You don't need metal maps either as this can be coped and mask painted as well. Same goes for dirt/dust/grime.
Here's a vid tut workflow overview that I did 2 years ago that might give you an idea of what I mean.