those lanterns are paper, the paper shouldn't have medium sized normal details, and the cuts would literally be a hole in the paper, not a weird scratch or chip.
thank you for sharing this. clearly just an experimental piece but damn if this doesn't hold some serious possibility. I have been playing around with NPR possibilities in UDK recently and I don't think I've seen anything quite this ingenious before.
while it really is nice, it definitely has some downsides when it comes to mipmapping, without custom code you cannot change the mipmaps in unreal as you can't load dds files, so from a certain distance its going to be massively blurry. there are quite a few cool hatching techniques even semi automatic ones but most of them will fail in udk due to the lack of mipmap abilities
one could buil a shader though to blend between differently scaled and detailled hatching textures but its going to be quite a heavy shader just for manipulating the "basic" shading, i did it once and decided to drop it as it was very heavy on instructions
thank you for sharing this. clearly just an experimental piece but damn if this doesn't hold some serious possibility. I have been playing around with NPR possibilities in UDK recently and I don't think I've seen anything quite this ingenious before.
Another 2d guy trying to learn 3d, eh? I'd say your 3d skills are perfectly fine!
Thank you Haikai. The fact it's that along the years, I never practiced 3d "seriously". As digital painting was always my primarly goal, I've studied a lot of things about Maya and Zbrush, but my knowledge was much more theoretical than practical. I don't feel confident to do 3d professionaly, understand? But this year I decided to focus on refine this, and I'm learning a lot. I myself didn't know that I could get such a result on Conan.. and there he is! This means that these studies are goign well.
while it really is nice, it definitely has some downsides when it comes to mipmapping, without custom code you cannot change the mipmaps in unreal as you can't load dds files, so from a certain distance its going to be massively blurry. there are quite a few cool hatching techniques even semi automatic ones but most of them will fail in udk due to the lack of mipmap abilities
one could buil a shader though to blend between differently scaled and detailled hatching textures but its going to be quite a heavy shader just for manipulating the "basic" shading, i did it once and decided to drop it as it was very heavy on instructions
answer from the man himself, i think you've forgotten more about working with shaders in UE3 than most people have ever learned
while we're on the subject what method did you end up going with for edge detection/cel shading in Airborn? I have been working through the "UDK gems" Sobel method, some of the stuff variously described on Eat3d (Oniram thank you for your tireless work and documentation there btw) and on the UDK forums and I am none too sure about the whole enterprise. I understand that basically any implementation is going to have its pros and cons, there being no one perfect catch-all solution, but with your previous experience I thought you might shed some light on it.
EDIT: further googling gets me breakdowns of various approaches you tried--so you ended up with a zbuffer-based edge detect as outlined on your post on GA here?
nice man really nice! been noticing you put out a killer sketch about every day or two, we gonna get to see any finished pieces from you? or just having too much fun with the speedies?
answer from the man himself, i think you've forgotten more about working with shaders in UE3 than most people have ever learned
while we're on the subject what method did you end up going with for edge detection/cel shading in Airborn? I have been working through the "UDK gems" Sobel method, some of the stuff variously described on Eat3d (Oniram thank you for your tireless work and documentation there btw) and on the UDK forums and I am none too sure about the whole enterprise. I understand that basically any implementation is going to have its pros and cons, there being no one perfect catch-all solution, but with your previous experience I thought you might shed some light on it.
EDIT: further googling gets me breakdowns of various approaches you tried--so you ended up with a zbuffer-based edge detect as outlined on your post on GA here?
yes but in the end we dropped all outlines because of a few reasons, I tested quite a bunch of methods, and the most interesting one isn't possible in UDK as you would draw actual edge meshes with a stroke texture applied (basicly you detect the edge, create polygonstrips aligned to the screen and map strokes on them) this is as far as i know the best possible solution.
The best working methods i had have been fresnel based and as you say zbuffer based. The problem with the first on is, that it is by design surfaceanglebased. So it is highly dependent on your mesh underneath, the denser and more homogenous your mesh is distributed the cleaner your outline will be, the best possible mesh for this would be a sphere/geosphere with equally sized triangles.
But as we all know, meshes fon't look like this, so the outline varies heavily on a mesh, theoretically you could use this effect for your favor and manipulate the outline by mesh manipulations, but to be honest its too complex to do in a somewhat reasonable timeframe.
And it all breaks on flat surfaces, Buildings are a no go, anything flat is a problem, best example would be Streetfighter4 Guiles hair on a specialmove, the top is all black because of the flat ange of that surface. Another issue with this method is scaleablity, theoreticly you could scale the thickness by distance, but it still doesn't cut it.
The second and better working method was zbuffer based, its pretty simple really you can do it in photoshop quickly, take a screenshot of your image, and one of your zbuffer (or render it, or just take a zbuffer from goole, doesn't matter) duplicate your zbuffer image, blur it by the amount of the thickness you would like your outline to be and put it on the layermode difference (mathwise this is blurred image - sourceimage) this is your outline. You can now take this and multiply it on top of your screenshot (or scene in unreal). The method is nice but has a few flaws.
First of all quite a few of instructions, especially if you want to "antialias" it, hardware antialiasing doesn't work as it is a posteffect so you have to blur (or sample) more often and layer it all together. With every row of pixels the instructioncount jumps higher and higher. But this was not the biggest flaw for us.
The biggest issue again had to do with the method beeing a posteffect.
It gets drawn last on the screen, obviously as it is a effect applied after everything else is drawn = POST effect.
With this in Mind a lot of problems appear, it doesn't work with fog, so you have to change the coloration over distance in the shader, of course one big plus for this technique is also that you can actually change the thickness and colour and visibility over distance.
So you can handle fog with more instructions, now comes the tricky part.
Particle are a HUGE issue, they get drawn before the post effects, so smoke or clouds are always behind posteffects even if the asset casting the outlines is behind a cloud or smoke or anything.
So Ships coming out of clouds, using clouds for cover, using smokegrenades as gameplay elements are impossible to do, you would als have outlines running in front of everything, the player would always know who was coming.
Another thing is, that the outline would always draw around everything (but softalpha assets as they don't get sorted the same way everything else is, so they don't appear in the zbuffer). So to avoid a few issues we thought of only drawing outlines around the character, and this is simples (at leas as far as we knew at this point) impossible in UDK.
So we dropped pretty much every possible dealbreake like outline and hatching.
Well by now we dropped pretty much everything besides the main idea, storyparts we already developed and the overall design. But we changed a lot, the style is different, the engine is different, some gameplay elements are different. But right now airborn is sleeping a very deep sleep, we have our second production running right now and we will only start to continue once everything is up and running and the office is crowded with people
Hi I am new in this community and so far is awesome >.<, here I share some of my work made completely in Maya 2011. Despite is finished I think i will improve some details.
The backstory: It's a magic temple under the sea.In the temple lies an object that brings magic to the surroundings, like animals and plants. I am a new in UV mapping and i did what i could :poly136:, but i liked the final result. I added some fog to the scene, projected an image trhough a spot light, bump to the ground, some lights into the jellyfishes and the plants. Please say me what u think about I just have 4 months of professional education in Maya, but I have been working lightly in it for like 1 year.
This is pretty much what I've been working on this weekend, a Hoverboard if you can't tell :P. Pretty much came from doodling/exploring in Max as I'm trying to learn that 3D package too. Well have a good one folks!
my Tacit Math reference folder is filling up nicely. I'm loving these last few and that flea thingy is awesomeness. Sort of Half life ish but more kick ass. I'd love to see a timelapse or some kind of quick workflow step by step
Its inspired me to try and do something similar (copy), expect it in my next WAYWO update.
Nice work all guys, i don't know why but the post I made is not here :S all right i post it again. I am new to these community and its awesome >.<. Here I post a work I did, despite its finished I plan to improve some details. It's a temple under the sea in which lies a magical artifact that gives magic power to the animals and plants in the surroundings.
Did a bit of texturing work, no specular yet. Thinking of rebaking the hair normals and adding some more interest to the back.
Thanks for the previous crits, I gave her some knuckles and fingernails, will definitely make a basket too, and an Uzi if there's time :>
Crits appreciated as usual!
while it really is nice, it definitely has some downsides when it comes to mipmapping, without custom code you cannot change the mipmaps in unreal as you can't load dds files, so from a certain distance its going to be massively blurry. there are quite a few cool hatching techniques even semi automatic ones but most of them will fail in udk due to the lack of mipmap abilities
one could buil a shader though to blend between differently scaled and detailled hatching textures but its going to be quite a heavy shader just for manipulating the "basic" shading, i did it once and decided to drop it as it was very heavy on instructions
Ahh good point, I didn't think about the mip maps, so you say if I create custom mip-map textures and switch them into the same shader network with if statements based off distance from screen it would be too heavy on instructions? I was just playing and am not planning on putting it into a game but I'm just currious.
Can't thank you enough, I don't know how long it would take to learn these lessons myself--could have been months or years before I ran into these issues myself and had to make some tough decisions.
While I am not entirely ruling out the possibility of using some kind of edge detection effect in my work, it seems pretty clear why it was bad fit for Airborn, and I especially appreciate explicit implications/limitations of employing such an effect. I have been toying with speccing a very flat visual style (largely inspired by Moebius/Geof Darrow, which seems to work well within the constraints established, but this is going to still take a lot of thought, planning, and experimentation.
Anyway, thanks again--I'll keep you updated with what I try out.
Replies
Another 2d guy trying to learn 3d, eh? I'd say your 3d skills are perfectly fine!
Thug Fortress 2?
thank you for sharing this. clearly just an experimental piece but damn if this doesn't hold some serious possibility. I have been playing around with NPR possibilities in UDK recently and I don't think I've seen anything quite this ingenious before.
nick: looks baws dude
got http://cghub.com/images/view/125489/ as my background atm, busted this dude out last night. Gotta get back to my lich ladeh for now.
one could buil a shader though to blend between differently scaled and detailled hatching textures but its going to be quite a heavy shader just for manipulating the "basic" shading, i did it once and decided to drop it as it was very heavy on instructions
Tried my first AO bake
didn't turn out that well. lots of splotches
have only a few ngons, project is WIP
not sure why this is happening. could anyone help? I tried at 512x2 and then 1024x2, made no difference
few of my guns finished recently...
thanks man.
yes all in marmoset!
Thank you Haikai. The fact it's that along the years, I never practiced 3d "seriously". As digital painting was always my primarly goal, I've studied a lot of things about Maya and Zbrush, but my knowledge was much more theoretical than practical. I don't feel confident to do 3d professionaly, understand? But this year I decided to focus on refine this, and I'm learning a lot. I myself didn't know that I could get such a result on Conan.. and there he is! This means that these studies are goign well.
Nick: thank you man!
nothankyou.jpg
id rather just see the model
\
answer from the man himself, i think you've forgotten more about working with shaders in UE3 than most people have ever learned
while we're on the subject what method did you end up going with for edge detection/cel shading in Airborn? I have been working through the "UDK gems" Sobel method, some of the stuff variously described on Eat3d (Oniram thank you for your tireless work and documentation there btw) and on the UDK forums and I am none too sure about the whole enterprise. I understand that basically any implementation is going to have its pros and cons, there being no one perfect catch-all solution, but with your previous experience I thought you might shed some light on it.
EDIT: further googling gets me breakdowns of various approaches you tried--so you ended up with a zbuffer-based edge detect as outlined on your post on GA here?
Nizza, really digging the turtle, can't wait to see how it will look when its complete.
This is my attempt at sculpting detail. It's a Garuda Mask.
nice man really nice! been noticing you put out a killer sketch about every day or two, we gonna get to see any finished pieces from you? or just having too much fun with the speedies?
-Woog
yes but in the end we dropped all outlines because of a few reasons, I tested quite a bunch of methods, and the most interesting one isn't possible in UDK as you would draw actual edge meshes with a stroke texture applied (basicly you detect the edge, create polygonstrips aligned to the screen and map strokes on them) this is as far as i know the best possible solution.
The best working methods i had have been fresnel based and as you say zbuffer based. The problem with the first on is, that it is by design surfaceanglebased. So it is highly dependent on your mesh underneath, the denser and more homogenous your mesh is distributed the cleaner your outline will be, the best possible mesh for this would be a sphere/geosphere with equally sized triangles.
But as we all know, meshes fon't look like this, so the outline varies heavily on a mesh, theoretically you could use this effect for your favor and manipulate the outline by mesh manipulations, but to be honest its too complex to do in a somewhat reasonable timeframe.
And it all breaks on flat surfaces, Buildings are a no go, anything flat is a problem, best example would be Streetfighter4 Guiles hair on a specialmove, the top is all black because of the flat ange of that surface. Another issue with this method is scaleablity, theoreticly you could scale the thickness by distance, but it still doesn't cut it.
The second and better working method was zbuffer based, its pretty simple really you can do it in photoshop quickly, take a screenshot of your image, and one of your zbuffer (or render it, or just take a zbuffer from goole, doesn't matter) duplicate your zbuffer image, blur it by the amount of the thickness you would like your outline to be and put it on the layermode difference (mathwise this is blurred image - sourceimage) this is your outline. You can now take this and multiply it on top of your screenshot (or scene in unreal). The method is nice but has a few flaws.
First of all quite a few of instructions, especially if you want to "antialias" it, hardware antialiasing doesn't work as it is a posteffect so you have to blur (or sample) more often and layer it all together. With every row of pixels the instructioncount jumps higher and higher. But this was not the biggest flaw for us.
The biggest issue again had to do with the method beeing a posteffect.
It gets drawn last on the screen, obviously as it is a effect applied after everything else is drawn = POST effect.
With this in Mind a lot of problems appear, it doesn't work with fog, so you have to change the coloration over distance in the shader, of course one big plus for this technique is also that you can actually change the thickness and colour and visibility over distance.
So you can handle fog with more instructions, now comes the tricky part.
Particle are a HUGE issue, they get drawn before the post effects, so smoke or clouds are always behind posteffects even if the asset casting the outlines is behind a cloud or smoke or anything.
So Ships coming out of clouds, using clouds for cover, using smokegrenades as gameplay elements are impossible to do, you would als have outlines running in front of everything, the player would always know who was coming.
Another thing is, that the outline would always draw around everything (but softalpha assets as they don't get sorted the same way everything else is, so they don't appear in the zbuffer). So to avoid a few issues we thought of only drawing outlines around the character, and this is simples (at leas as far as we knew at this point) impossible in UDK.
So we dropped pretty much every possible dealbreake like outline and hatching.
Well by now we dropped pretty much everything besides the main idea, storyparts we already developed and the overall design. But we changed a lot, the style is different, the engine is different, some gameplay elements are different. But right now airborn is sleeping a very deep sleep, we have our second production running right now and we will only start to continue once everything is up and running and the office is crowded with people
THE WALL OF TEXT ENDS HERE
:poly142:
ps: tacit you rock
Very very cool !
The backstory: It's a magic temple under the sea.In the temple lies an object that brings magic to the surroundings, like animals and plants. I am a new in UV mapping and i did what i could :poly136:, but i liked the final result. I added some fog to the scene, projected an image trhough a spot light, bump to the ground, some lights into the jellyfishes and the plants. Please say me what u think about I just have 4 months of professional education in Maya, but I have been working lightly in it for like 1 year.
goodness, i MUST learn automotive modelling. street scenes need them, and im not the best at it lol
nice work there
Le ruin~
4164 tris' Just a piss about project while i dig outa a rut.
my Tacit Math reference folder is filling up nicely. I'm loving these last few and that flea thingy is awesomeness. Sort of Half life ish but more kick ass. I'd love to see a timelapse or some kind of quick workflow step by step
Its inspired me to try and do something similar (copy), expect it in my next WAYWO update.
Keep them coming.
p.s. This should be on a T shirt
Banged out some thumbs and then refined my fav.
Joust Mech. Will block it out next.
Thanks for the previous crits, I gave her some knuckles and fingernails, will definitely make a basket too, and an Uzi if there's time :>
Crits appreciated as usual!
little update of my croc thing, just a very dirty quick ingame test. Rendered with cryengine3..
Ahh good point, I didn't think about the mip maps, so you say if I create custom mip-map textures and switch them into the same shader network with if statements based off distance from screen it would be too heavy on instructions? I was just playing and am not planning on putting it into a game but I'm just currious.
future-fiction - I was hoping I would see her again.The face looks amazing.All the soft color blends.Very anime esque.
tacit math - Are those fullblown sketches or sculpt/2d??
Polycount is so beast.
Slowly carrying on! There are a couple of control loop issues but I will fix them in good time!
keep it up!
Wheelchair HP... about 90% done.
Fucking Beast Bro~!
first AO bake, woot. finally got it to work
Fleshin out the internals.
I second Striker's comment. The stiff neck kinda kills the motion.
I love it tho. Refreshing.
I have to agree, theres some very good work here - but she looks like she just caught whiff of a very eggy fart.
Alberto Rdrgz - Thanks man.More on the way.
Dan! - Thank ya, been missing your art.What have you been up to lately?
hawken - YOUR SOOO RIGHT THOUGH LololOL???I will try to keep the farts to a minimum next time man.
Can't thank you enough, I don't know how long it would take to learn these lessons myself--could have been months or years before I ran into these issues myself and had to make some tough decisions.
While I am not entirely ruling out the possibility of using some kind of edge detection effect in my work, it seems pretty clear why it was bad fit for Airborn, and I especially appreciate explicit implications/limitations of employing such an effect. I have been toying with speccing a very flat visual style (largely inspired by Moebius/Geof Darrow, which seems to work well within the constraints established, but this is going to still take a lot of thought, planning, and experimentation.
Anyway, thanks again--I'll keep you updated with what I try out.