I think this sentiment has been around for the past decade+ as geometry has not been a GPU bottleneck/performance issue for a looong time.
Just give yourself enough fidelity to look good enough, and within reason and sensibility. There really isn't a goal/benchmark of what triangle limit to aim for, it just so happens to land around this arbitrary number for many factors. (subject matter has lots of armor pieces, super organic, plus lots of hair cards = inherent high triangle count)
e.g. don't go nuts and author several iterations of subdivisons "just cause", effectively skyrocketing your geo to 400k triangles or whatever.
but also don't skimp out on areas that could use more silhouette/loops/fidelity - we don't need to second guess if we can "afford" it anymore.
Something like 250 000 would seem like a decent number to me. Those numbers have been run on the PS4 too for cinematics, should be trouble-free to go with on a modern PC. High res enough for detailed closeups and if need be just do a lower level for the actual third person game camera view with like 2/3rds of that polycount. Assuming your character isn't bald I'd put aside around 50 000 for hair if done with polygons and not the groom stuff that is supported in UE.
Pretty sure you could use more these days - but obviously the higher the budget the more complex the details the longer it all will take to create.
so we're working on an indie project using third person and cinematic close-ups in Unity. note, our typific art style allows us fewer polycounts. i was not given a poly target by our engineers but approached just using common sense practices and these are the counts they ended up at. we later created an uprez'd head for cinematics that about doubled its counts from the neck up.
our main character variant A is 11k polies and variant B is 13k. their weapon is 1k and additional accessories are at 1k. the uprez'd cinematic bust is at 5k.
our enemies clock in at around 5-15k.
(for tricount just double these numbers)
our programmer had informed me i could double these counts with no problem. i'm like that old guy that grew up during the depression that uses the same tea bag for 6 months, only it's polycount.
dunno what the context is but if you are figuring out targets at the same time you are doing the work... mentally prepare yourself to do it again. Like five more times.
@Kanni3d I hear you man! Yeah - not trying to go nuts but was also trying to push it from what I'm used to. Been doing low-res work for work for a long time and wanted to get away from that and try something a bit hi res. Thankfully no hair cards on this one! Wearing helmet.
@thomasp wow, 250k sounds awesome! But that probably will be a lot to manage for me, not used to dealing with such hi res. But it's awesome that you think it could go that high. I think, based on what I read here, I will aim for about 50-60k which is in the realm of something I think I can handle comfortably when it comes to the rigging and UVs.
@oglu its just your standard space marine kind of guy. think master chief or doom guy.
@killnpc that sounds very manageable, but I'm going for a bit higher res. I hear you on polycounts though, I normally favor lower, but I've been doing that for work for so long wanna try something else.
@Alex_J thanks for you fine words of encouragement XD
When I did my numbers I just went with your original requirements: AAA, UE5, PC, cinematic quality and 2020's standards. Making a character to those standards would keep you busy for months for sure. Not necessarily fun to do all that. And yes you could go that high (in my experience).
Every game is unique. There are too many factors that you need to consider. And the bottleneck isn't the tri count anymore, as mentioned before. Shading, textures, game logic code, particle effects in the scene, and so on.
@thomasp woo! nice! yeah that range seems good to me too. One thing that keeps me from going higher is that I use 3d Max to rig, and it's extremely painful to rig in max with hi poly things (at least for me) In fact, I've been thinking of using primarily weight maps based on the UVs rather than go the traditional painful way of painting weights manually on the model.
@dread_companion One way to speed up skinning in Max that I used years back was to model a really low polygon mesh, apply the Skin modifier and tweak to my liking, then skin-wrap (modifier) a higher resolution model to this mesh and use 'Convert to Skin' to turn the skin wrap into a new skin modifier. That usually left only the finer points for tweaking.
I usually did something like two layers of skinning proxies - a low one that only contained the basic limbs, head and body skinning and a medium resolution one based on it that contained fingers, etc. Tweaked each of these till it was fine before converting it, then skin-wrapped my actual game mesh to it and did the same thing there.
Replies
how long is a piece of string?
singleplayer? multiplayer? what kind of AAA game? first person? third person? What kind of genre? How far are you away from said character?
Hi Neox.. lol
Yeah i was probably too non specific.
Third person AAA, but there will be close ups in cinematics. Example, Mass Effect.
Genre, Sci-Fi
I'm doing the sculpt retopologizing now... and that's where the question came up
That also depends on the design of the char. Show us some images.
I think this sentiment has been around for the past decade+ as geometry has not been a GPU bottleneck/performance issue for a looong time.
Just give yourself enough fidelity to look good enough, and within reason and sensibility. There really isn't a goal/benchmark of what triangle limit to aim for, it just so happens to land around this arbitrary number for many factors. (subject matter has lots of armor pieces, super organic, plus lots of hair cards = inherent high triangle count)
e.g. don't go nuts and author several iterations of subdivisons "just cause", effectively skyrocketing your geo to 400k triangles or whatever.
but also don't skimp out on areas that could use more silhouette/loops/fidelity - we don't need to second guess if we can "afford" it anymore.
Something like 250 000 would seem like a decent number to me. Those numbers have been run on the PS4 too for cinematics, should be trouble-free to go with on a modern PC. High res enough for detailed closeups and if need be just do a lower level for the actual third person game camera view with like 2/3rds of that polycount. Assuming your character isn't bald I'd put aside around 50 000 for hair if done with polygons and not the groom stuff that is supported in UE.
Pretty sure you could use more these days - but obviously the higher the budget the more complex the details the longer it all will take to create.
so we're working on an indie project using third person and cinematic close-ups in Unity. note, our typific art style allows us fewer polycounts. i was not given a poly target by our engineers but approached just using common sense practices and these are the counts they ended up at. we later created an uprez'd head for cinematics that about doubled its counts from the neck up.
our main character variant A is 11k polies and variant B is 13k. their weapon is 1k and additional accessories are at 1k. the uprez'd cinematic bust is at 5k.
our enemies clock in at around 5-15k.
(for tricount just double these numbers)
our programmer had informed me i could double these counts with no problem. i'm like that old guy that grew up during the depression that uses the same tea bag for 6 months, only it's polycount.
dunno what the context is but if you are figuring out targets at the same time you are doing the work... mentally prepare yourself to do it again. Like five more times.
@Kanni3d I hear you man! Yeah - not trying to go nuts but was also trying to push it from what I'm used to. Been doing low-res work for work for a long time and wanted to get away from that and try something a bit hi res. Thankfully no hair cards on this one! Wearing helmet.
@thomasp wow, 250k sounds awesome! But that probably will be a lot to manage for me, not used to dealing with such hi res. But it's awesome that you think it could go that high. I think, based on what I read here, I will aim for about 50-60k which is in the realm of something I think I can handle comfortably when it comes to the rigging and UVs.
@oglu its just your standard space marine kind of guy. think master chief or doom guy.
@killnpc that sounds very manageable, but I'm going for a bit higher res. I hear you on polycounts though, I normally favor lower, but I've been doing that for work for so long wanna try something else.
@Alex_J thanks for you fine words of encouragement XD
Thanks y'all!! :)
With or without helmet? A head with hair does eat a lot of poly. You see there is no easy answer without knowing what you try to create.
Sounds reasonable to go with 50-60k!
When I did my numbers I just went with your original requirements: AAA, UE5, PC, cinematic quality and 2020's standards. Making a character to those standards would keep you busy for months for sure. Not necessarily fun to do all that. And yes you could go that high (in my experience).
My advice is, make a prototype and find out :)
Every game is unique. There are too many factors that you need to consider. And the bottleneck isn't the tri count anymore, as mentioned before. Shading, textures, game logic code, particle effects in the scene, and so on.
@thomasp woo! nice! yeah that range seems good to me too. One thing that keeps me from going higher is that I use 3d Max to rig, and it's extremely painful to rig in max with hi poly things (at least for me) In fact, I've been thinking of using primarily weight maps based on the UVs rather than go the traditional painful way of painting weights manually on the model.
@Glear indeed my friend... indeed.
@Tiles that's actually what I'm doing!! So... we'll see. I'll keep you posted. :)
@dread_companion One way to speed up skinning in Max that I used years back was to model a really low polygon mesh, apply the Skin modifier and tweak to my liking, then skin-wrap (modifier) a higher resolution model to this mesh and use 'Convert to Skin' to turn the skin wrap into a new skin modifier. That usually left only the finer points for tweaking.
I usually did something like two layers of skinning proxies - a low one that only contained the basic limbs, head and body skinning and a medium resolution one based on it that contained fingers, etc. Tweaked each of these till it was fine before converting it, then skin-wrapped my actual game mesh to it and did the same thing there.
https://knowledge.autodesk.com/support/3ds-max/learn-explore/caas/CloudHelp/cloudhelp/2022/ENU/3DSMax-Modifiers/files/GUID-1B3C4662-8C42-4F02-BD1B-204CD491166B-htm.html
Helps a lot when the proxies use the same basic topology as the game mesh to make the most of this.
@thomasp oh, wow 😮 thanks for the tip!! I'm definitely looking into this!