It just annoys me, that certain people keep advocating the slowest possible methods to newbies, instead of investigating why I'm saying that's not the way to do things.
I fixed every screwed-up normal in CA's meshes in less than an hour, using only UpSpring. I've offered, over and over again, to fix any model that these Slow Food guys seem to think that I can't get fixed.
Seriously- show me the money, guys. Show me some models from your games that cannot be fixed up, fast, using UpSpring. Send it to me with the verts all welded, basically fubar. I will make it look right, and if UpSpring isn't the way, then if nothing else we can see an example of where this stuff fails, and pass THAT on to newbies or lurkers.
I've even told you folks, several times, exactly the kind of situation you'd need to create. It's certainly possible, if you do it on purpose.
But in actual production... it's rare! That's my whole point here! Importing models a Piece at a time is very, very inefficient, when you look at it this way- it's like saying that since people occasionally get hit by lightning, that you'll never walk outside.
I really don't think you'll be able to, unless you're regularly trying to treat cuboids as "round", which you shouldn't be doing in the first place. I've seen your four-sided gun barrels, man. They still look four-sided. The minimum before it starts coming out halfway round is five sides. To get enough tesselation for it to be really convincing, you need at least 8. I feel like you guys have learned all the wrong things about low-poly modeling, if you're agonizing over polycounts on gun barrels that are teensy by any modern standards.
Meh. I forget where I first read about the "rule of 89", but it certainly well applies to Spring modeling, or modeling in any environment that requires shaders. If you need hard edges on something with angles greater than 89 degrees, you're doing it wrong. 3D engines can fake a lot of stuff, but turning concave into convex is not something they can do... and asking them to is a mistake. If that's how you modeled it, live with the blockiness, on most stuff like that it doesn't matter in terms of what players will notice anyhow...
Geeze, cool off. Argh is right, for 90% of cases the autogenerated smoothing does it just fine... but Smoth is quite obviously making his Gundam units as a labour of love and fandom, and they look exceptionally good for their low polycount. So obviously he's going to put a little extra TLC into the smoothing and do it _properly_ rather than letting the autogenerator do it for him.
You two are never going to agree on this - you simply have different standards for Good Enough - obviously I agree with Argh, but you can't fault Smoth for being a perfectionist with his _own_ work, and for telling others how to achieve similar perfection if they should choose to do so.
But imho, Argh is right about this thread. There are simply too many units in TA (and most other Spring games) to apply Smoth's level of quality to every unit unless you're willing to commit to Smoth's level of labour - that is to say, units that look gorgeous despite low-poly construction. And we rarely see the units up-close-enough to need Smoth's level of quality, particularly when their hyper-low-polycount means they're hyper-spammable. Smoth's designs would be farking awesome for a DS game where you see the super-low-poly units up-close.
But either way, I'm not going to call him wrong for wanting to pour that love into his units, or for telling others how to do so.
A method that destroys a significant portion of the model's artistic data is being presented as a standard - even suggested to those who specifically wish to preserve their painstakingly created smoothing groups. Shall we add a button to upspring that deletes every 10th vertex, and present that to new modellers as a good idea as well? Or how about removing the green channel from all textures? After all, working with all those vertices and an obviously redundant *three* primary colours takes up a lot of the artist's time; streamlining those bottlenecks in the process will speed up the process greatly and will obviously be the correct solution 90% of the time.
Verts are either welded, or they aren't. All a smoothing group does (in S3O, mind you, it's a lot more complex in 3DS, which is where that term comes from- "smoothing group" is a subdivision modeling thing, IDK how it got perverted around here... oh yeah, Wings is a sub-div modeler too, that explains a lot...) is determine which verts are welded.
Do you guys not understand welding? Is that the basic problem here?
Well, basically... at an intersection of any two triangles, you have two vertices... or you have one. If you have two, they're not welded, and if two triangles have two points in common, and they aren't welded, then that angle registers as "hard".
Welding reduces vert count- it literally makes the data load smaller, and it's an important part of efficient modeling.
But Spring handles welded verts in a specific way. In Spring, that gets translated as "round" by the shaders- they treat the welded verts as one continual surface, basically. I don't pretend to know the full details about that- it's either due to gl.smooth or it's because the vertex shaders see welded verts as one thing, and thus two adjoining welded triangles get treated as one surface when being tesselated. I don't know the precise answer to that- that's a question for jK, Kloot, jcnossen et al.
All that UpSpring is doing, when you use "recalculate vertex normals, 3DO style", is it either welds the verts... or it doesn't. And it's using a breaking angle to do this- the angular difference between the normal of the triangle vs the ones around it, which is expressed as a vector. I'm sure that jcnossen probably refers to that in different technical terms, but that's essentially all that's happening here.
So... all you're doing is adjusting the maximum breaking angle. Start low, at about 45 degrees, then work up until your rounded surfaces are rounded in appearance in UpSpring... then you're done.
It's very easy, it's very fast, and it's worked on all but ONE MODEL I've encountered, including a lot of stuff I didn't model. That one model was mainly a mistake- I included a surface with very low breaking angles that I wanted hard edges on, instead of exporting those triangles as a separate object, which is what I did to fix it.
It's not hard, people. It's not "bad practice". It's not what you should be teaching newbies is evil. And if you're actually having trouble understanding how to do it, or why it works... that's not my problem.
Last edited by Argh on 08 May 2009, 01:43, edited 1 time in total.
please condescend more mr. teamcolors are an additive blend, mr. salmon pink is red.
*edits* on second thought I don't think i am going to talk to you any more. Please stay the fuck out of my threads and I will leave you to stroke your dick to shallow efforts that make up your "groundbreaking" crap.
I don't want to burn the bridge with you, I want to blow it up and piss on the ashes. FUCK OFF
Users browsing this forum: No registered users and 1 guest
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot post attachments in this forum