so basicly it is a 16bit displacement map that you use as a 8bit bumpmap?
and if thats what zbrush produces as a normalmap, somethig is definitely off, did you try xnormal before? good luck with using this map from now on, might work for your portfolio but will kick you in the ass once you are in production
Bump maps are just an indirect way of creating normal maps. Bump maps don't contain height information, just a very sloppy approximation. Bump maps need to be multi-sampled to produce a normal, which effectively blurs the result.
Not sure what you mean by this. Normal map normals are per-pixel, meaning true-rez.
the gradient/ height difference is generated between the pixels on a bump map so this can give the impression of higher res/ sharper edge lines....depends on the filtering employed i think...
and how close up you get to the map...IE a high res bump at a distance looks great but up close a low res wont give as good a result as a normal map
Correct me if I'm wrong though, but wont most engines nowadays convert bumpmaps into normalmaps in the end anyway?
That is, considering the fact that the end result texels will have a normal direction, which is either generated from the bumpmap, or just accessed directly from a normalmap.
The interpolation between the two pixels are the same as interpolating between two diffuse pixels.
U can change the pixel blending to an extend where you end up with an more pixelated result. But you get sharper lines making you think the texture is more highres. Normalmaps are usually blend "physically correct" what is good but you need more then 9 pixel to define a sharp bump (e.g. a screw or something)
So when I'm talking about resolution I'm not talking about technical stuff here in contrary to what you are referring to with normal-per-pixel-resolution per.
Correct me if I'm wrong though, but wont most engines nowadays convert bumpmaps into normalmaps in the end anyway?
That is, considering the fact that the end result texels will have a normal direction, which is either generated from the bumpmap, or just accessed directly from a normalmap.
The interpolation between the two pixels are the same as interpolating between two diffuse pixels.
Yes, I'd also say majority do that conversion, just like bumpmap NV photoshop filter or other stuff. You need a normal in the end, and sampling neighbors to generate it from height differences is something you don't want to do at runtime, you just want to sample the texture once and get the normal (or gradient) directly.
in the really old days, like when Per was in his fortys, "bump mapping" was "emboss bump mapping", as there was no dot-product (normal/direction angle) instuction in hardware yet. http://www.tweak3d.net/articles/bumpmapping/ (that article is from 10 years ago hehe, so it basically ended with an outlook on per-pixel lighting hehe)
Bump mapping is no secret but that's not what I'm talking about. It looks like the "special map" has some kind of wanted or unwanted dither in it, making the whole thing look more highres than the nm
+ it's simpler to sharpen the black n white bump map compared to the nm.
And this is totally on topic bc I guess that's what's going on
This is about the feeling of higher resolution not technical a higher resolution
But I have to admit that my first sentence was not clear enough defined to make it per proof.
I can completely understand the fact that this will never REPLACE a normal map, that's just not possible, however, when used in conjunction with an already made normal map from the same object, perhaps multiplied over it in Photoshop, this thing really adds some nice touches that don't seem to come through all the way in ZBrush's Normals, if you use ZBrush's normal creation tools etc..
Yes, I'm aware these maps are absolutely terrible lol, I made them very quickly without any intentions of making them look excellent. I just want the concept of whats going on here to be gotten acrossed.
Yes, I think xNormal is amazing However, it was fun to see what could be done with this Gray Map, considering all the things that could be done with it since it is a grayscale.
I actually took this map and converted it with xNormals Photoshop Filter tools, and laid it on top of a normal map from the same geometry of which I created in ZBrush, and it really did give me some amazing results.
OF all things, I'd say that this map could be purposed as a GUIDELINE to how good a NORMAL could be if made correctly for that specific object. I used it to really tweak my normals to look like the exact same thing / better as those "Special" versions, but instead as a beautifully working Normal map.
I suggest you guys don't write it off instantly as useless and play around with it. No harm can come of it really.
This is sort of a silly thread, and there are a lot of misconceptions here. A bump map that is converted to a normal, has the exact same limitations as a normal. One can not be "sharper" than the other, its impossible, its limited the the bounds of the pixels contained in the image.
If you want more sharpness in your result, you have a few choices
1. Make your model sharper
2. Increase your texture resolution
3. Sharpen your normals in photoshop. Contrary to popular belief this will not explode your computer.
This whole thing is trying to throw some random shit together to make up for the fact that the artist isn't capable of getting a good result, as very evident by the example images. This is either an artistic or a technical problem, and not something that can be fixed by throwing random shit on your textures and hoping for the best.
OF all things, I'd say that this map could be purposed as a GUIDELINE to how good a NORMAL could be if made correctly for that specific object. I used it to really tweak my normals to look like the exact same thing / better as those "Special" versions, but instead as a beautifully working Normal map.
A normal map bake is the direct result of result of the content you use to create it, the most straight forward way to create a good normal map is to create good content. Tweaking your normalmap in photoshop without really understanding how a normal map works is definately not recommended, as generally your normal map is baked to match your lowpoly's mesh normals, and as soon as you start doing crazy stuff to it, you will introduce all sorts of artifacts and errors.
I'm with EQ. There's not really any good use for this. Nobody uses bump for realtime shaders, and you're much better off just making a proper normal map in the first place than going through some bizarre highpoly->heightmap->filters->normals process. One of the main reasons his example "special map" looks higher resolution is simply because his convoluted process introduced noisy accuracy errors that the Photoshop filters exaggerated. It's like throwing a per-pixel noise image into Crazybump and layering the result over your highpoly normal. I'm not familiar with using ZB to generate normals but if those two textures are in fact the same resolution, either his sculpt was mushy or he did a bad job on baking the normals.
It can be very very very useful for shading/texturing, and more when the depth is 16 bit color. For me it's another greyscale map, so very useful as the "curvature maps" that we can get with Zbrush or other kind of grey scale maps.
Arguing about if it's useful it's a nonsense, play with it and find its utility. I can say it has its potential.
in my earlier post i was talking about oldskool emboss bump mapping...and the one advantage i remember them having....
curious if any engines that used detail grey scale "bump" maps ( i remember a few being talked about a few years ago but had no experience with them) are really that if as you say they converted to normals that would seem expensive...
interesting but slightly pointless conversation as its application in realtime is mostly historic
I just don't get the whole "no one knows how its called, not even pixo" thing because step 6 says what its called, displacement map and of course its usefull and of course it can be used to enhance things like the textures
in my earlier post i was talking about oldskool emboss bump mapping...and the one advantage i remember them having....
curious if any engines that used detail grey scale "bump" maps ( i remember a few being talked about a few years ago but had no experience with them) are really that if as you say they converted to normals that would seem expensive...
interesting but slightly pointless conversation as its application in realtime is mostly historic
Yeah, Doom3 is a good example, you can input a grayscale bump, but it doesnt actually use this to render, it converts it to normals and combines the two(bump+generated) on load(or compiles it into a compressed texture or something once at re-uses that).
I guess bc its funky stupid putting things on and of and doing this and that and eating bats and coocing snakes and if you HAVE the spirit you will be able to procue the special special map
it's not the first time we see something like this :poly142:
This is sort of a silly thread, and there are a lot of misconceptions here. A bump map that is converted to a normal, has the exact same limitations as a normal. One can not be "sharper" than the other, its impossible, its limited the the bounds of the pixels contained in the image.
If you want more sharpness in your result, you have a few choices
1. Make your model sharper
2. Increase your texture resolution
3. Sharpen your normals in photoshop. Contrary to popular belief this will not explode your computer.
This whole thing is trying to throw some random shit together to make up for the fact that the artist isn't capable of getting a good result, as very evident by the example images. This is either an artistic or a technical problem, and not something that can be fixed by throwing random shit on your textures and hoping for the best.
Hey guys,
I thought it was something interesting that could be played around with, that's all. Many of you are acting like I'm creating a new religion or something...chill the f**k out for a second and just look at it for what it is, maybe? I'm "Marketing" this, you say? Okay, sure thing, business man.
Like I said, use it at your own discretion. I'm not trying to prove anyone better or worse or right or wrong here (of which many of you seem to love to do in this thread).
As an ARTIST, you guys should be excited and explorative when something like this comes along, and not cynical and close-minded. Maybe some of you still have something yet to learn that goes beyond the technicalities....actually, that's quite obvious the way some of you are acting.
But anyways, back on topic, it's interesting that this kind of map can be produced out of Zbrush. Is there anyone who's played around with it at all?
I just don't get the whole "no one knows how its called, not even pixo" thing because step 6 says what its called, displacement map and of course its usefull and of course it can be used to enhance things like the textures
Did you read the thread I linked to, at all?
RASTAMAN found this stuff, and he contacted Pixo and that's what he said they replied with.
What i don't get is, why the detail is more high than in a normal map. I mean, if you create a normal map FROM a high map then you need 4pixels of the high map the calculate one pixel of the normal map. That's not good - OK!
But when i render the normal map directly from the high poly model. Shouldn't i get exactly the same amount of detail like with this special map / displacement map?
As an ARTIST, you guys should be excited and explorative when something like this comes along, and not cynical and close-minded. Maybe some of you still have something yet to learn that goes beyond the technicalities....actually, that's quite obvious the way some of you are acting.
I think people are more willing to listen to someone who actually knows how things works, and I mean ACTUALLY works, like normalmaps, bumpmaps, displacements etc, but yours has been mostly guessing in the name of "exploration". You you clearly noted that the specia.. bumpmap looks way better than a normalmap when in reality you really should get better at baking and handling normalmaps, your second example even shows that your examples aren't valid anymore, since the normalmap is all stretched all over it.
But sure, experimenting is fun and all, but in the end, without raw knowledge, you'll just end up doing things you dont understand, you'll just end up misleading other people who also don't know how things work. Those mislead people just end up here and cause a headache for everyone.
Technicalities are important since computers will always follow rules, it's their very nature, you can't just randomly break them.
I shall show you this entertaining thread that popped up on gamedev one day:
What i don't get is, why the detail is more high than in a normal map. I mean, if you create a normal map FROM a high map then you need 4pixels of the high map the calculate one pixel of the normal map. That's not good - OK!
But when i render the normal map directly from the high poly model. Shouldn't i get exactly the same amount of detail like with this special map / displacement map?
If the maps are the same size they should represent the same amount of texels on the model surface, so they should have equal fidelity, but the normalmap baked from a highpoly is exactly that, an exact representation of how the highpoly looks, with that amount of texture fidelity.
You only need a pixel from a heightmap to create some normalmap differences, in the end it's all about how the heightmap is chosen to be intepreted.
I thought it was something interesting that could be played around with, that's all. Many of you are acting like I'm creating a new religion or something...chill the f**k out for a second and just look at it for what it is, maybe? I'm "Marketing" this, you say? Okay, sure thing, business man.
Like I said, use it at your own discretion. I'm not trying to prove anyone better or worse or right or wrong here (of which many of you seem to love to do in this thread).
As an ARTIST, you guys should be excited and explorative when something like this comes along, and not cynical and close-minded. Maybe some of you still have something yet to learn that goes beyond the technicalities....actually, that's quite obvious the way some of you are acting.
But anyways, back on topic, it's interesting that this kind of map can be produced out of Zbrush. Is there anyone who's played around with it at all?
no i didn't read the extra thread but reading your instuctions tells me its nothing new its just the way you did displacements and stuff in zbrush before they introduced their new tool for map creation - i don't even remember its name anymore because it produced bad flat normals when i last used it.
I think people are more willing to listen to someone who actually knows how things works, and I mean ACTUALLY works, like normalmaps, bumpmaps, displacements etc, but yours has been mostly guessing in the name of "exploration". You you clearly noted that the specia.. bumpmap looks way better than a normalmap when in reality you really should get better at baking and handling normalmaps, your second example even shows that your examples aren't valid anymore, since the normalmap is all stretched all over it.
But sure, experimenting is fun and all, but in the end, without raw knowledge, you'll just end up doing things you dont understand.
Technicalities are important since computers will always follow rules, it's their very nature, you can't just randomly break them.
I shall show you this entertaining thread that popped up on gamedev one day:
I definitely am a learning artist, like many here. Yes, I said the normal map, without any tweaks, looked less good than the special map, without any tweaks. Without any tweaks. It was a raw comparison, visually. I also stated the examples maps I made were done quickly and were crap....I was just getting the concept across about what I thought was going on. And, who said I was talking about using this in just video-games or in production animations etc? If it looks good in a still shot / render, it looks good in a still shot /render, no?
Yeah, computers follow rules, and then you find glitches like this. What's your point, man?
Why am I arguing? I came to show that this exists. If you don't like my opinions on it, that doesn't really matter. I'm willing to learn more, but telling ME personally that "I don't really understand technicalities" and basically telling me no-one should listen to me is absolutely besides the point. When did I ever say "YOU MUST ALWAYS FROM NOW ON USE THIS MAP. USE IT NOW. I COMMAND IT" lol....
So, what I'm saying, is stop the *I'm better than you* routine and just get on with our lives here lol...I simply am having to defend myself from...I don't know what, exactly? Smarty pants people who are trying to talk down to me? I'm not sure really...
I simply thought this would be something interesting the community could play around with?
The problem is, these forums are filled with people who work with surface-defining shaders and maps all day long as a living, and many of them know their way around it quite well.
There's also the people who would overlay a normalmap unto itself since they heard it was a good trick to enhance the looks of it!, that kind of blind experimentation mentality isn't pretty, since it never gets anywhere.
I wasn't refering to the quality or anything about your examples, I was refering mostly to the fact that the pillar clearly isn't mapped the same on both examples, and it would help if you showed the actual highpoly you baked from.
We're not doing the "better than you" routine, but you do have to accept, this is polycount, good things comes out from listening to the people who have experience with these kind of things on these boards.
People in this thread have said things along the lines of "this isnt really a good idea, big waste of time.". Its important to know that this is not only not a good idea, its a very very bad idea.
If you go through the trouble of modeling high res geometry you better be baking it directly to a normal map. Rendered normal maps define the exact difference between the low res geometry and the high res geometry which is needed for correct rendering. If you were to try to render a displacement map from these 2 objects and then try to convert it to a normal map you will end up with horribly inaccurate results, which will result in lighting errors/seams/artifacts in your renders.
The problem is, these forums are filled with people who work with surface-defining shaders and maps all day long as a living, and many of them know their way around it quite well.
There's also the people who would overlay a normalmap unto itself since they heard it was a good trick to enhance the looks of it!, that kind of blind experimentation mentality isn't pretty, since it never gets anywhere.
I wasn't referring to the quality or anything about your examples, I was refering mostly to the fact that the pillar clearly isn't mapped the same on both examples, and it would help if you showed the actual highpoly you baked from.
We're not doing the "better than you" routine, but you do have to accept, this is polycount, good things comes out from listening to the people who have experience with these kind of things on these boards.
I completely agree. But, blind experimentation? You may know much more about shaders and what-not, but can you really be an artist if you do not explore? My experimentation was not blind, just hopeful. I do know what a normal map is and how it works. Again, I never said this has to be exclusive to video-games.
It has also been made very apparent to me that mixing images and maps in photoshop is NOT something frowned upon. Why would you say...
"There's also the people who would overlay a normalmap unto itself since they heard it was a good trick to enhance the looks of it!, that kind of blind experimentation mentality isn't pretty, since it never gets anywhere."
I don't agree with you there...
I know MANY professionals currently at game studios, of whom used to be professors of mine, who would do that kind of thing regularly and suggested that it was something that you needed to TWEAK / EXPERIMENT with, and that great results could come of it. In fact, tweaking is experimentation in itself, is it not?
My whole point is leading to the fact that experimentation is good, and in me bringing this over to polycount I hoped to see some people experiment with it to see what cool things could be done. By saying something along the lines of "Blind experimentation doesn't get anywhere," not only are you putting aside the fact that this kind of thing could possibly be useful somewhere, you are being oblivious to the fact that experimentation IS a professional way to go about things, and by those means, it should be encouraged.
I will also say that I would love to listen to what professionals have to say here on these forums, pertaining to this kind of thing. Perhaps someone who does not have to talk down to a person whilst teaching them a thing or two?
People in this thread have said things along the lines of "this isnt really a good idea, big waste of time.". Its important to know that this is not only not a good idea, its a very very bad idea.
If you go through the trouble of modeling high res geometry you better be baking it directly to a normal map. Rendered normal maps define the exact difference between the low res geometry and the high res geometry which is needed for correct rendering. If you were to try to render a displacement map from these 2 objects and then try to convert it to a normal map you will end up with horribly inaccurate results, which will result in lighting errors/seams/artifacts in your renders.
I completely agree there, concerning normal maps and such. However, is this map really a displacement map? I don't work with disp maps too often, because they dont work in real-time. They do look different do they not? Or am I wrong? Also, it doesn't HAVE to be converted to a normal map. Could it work for anything else? That's what I'm also interested in...
But it doesn't HAVE to be converted to a normal map
but that is the main point here...
People have used various greyscale maps (curvature, displacement maps, dirt...) to enhance their textures, is nothing new and common practice.
what you suggested (the really bad normalmap bake picture in your first post) is that normalmaps are inferior to this "special map" (displacement -> normalmapfilter) workflow. And that is very very bad, as stated by numerous people here.
And because that is your main point (first post) and the main gist of this "marketing" like picture (really crappy normalmap bake vs what a normalmap normally looks like "special map"). That is what upsets people. Not the fact that you experiment with greyscale or whatnot maps.
I completely agree. But, blind experimentation? You may know much more about shaders and what-not, but can you really be an artist if you do not explore? My experimentation was not blind, just hopeful. I do know what a normal map is and how it works. Again, I never said this has to be exclusive to video-games.
Then you do know that a normalmap that is baked from a highpoly will in the best scenario be a perfect replica of the highpoly (with an exception of the silhuette) so in most cases your highpoly will be at fault.
And these normals are also baked according to the tangents of your lowpoly, and this is important to know, because if you alter these in a way such as overlaying the entire normalmap on itself, you're pushing them out of alignment, so the normalmap will no longer be correct.
People have used various greyscale maps (curvature, displacement maps, dirt...) to enhance their textures, is nothing new and common practice.
what you suggested (the really bad normalmap bake picture in your first post) is that normalmaps are inferior to this "special map" (displacement -> normalmapfilter) workflow. And that is very very bad, as stated by numerous people here.
Let me say now, that when I stated that I did not know what the "Special" map was all about, it was still just starting to get around to quite a few people at ZBrushCentral and myself. It's definitely old now in comparison to the new knowledge gained from this place. The fact is, in Maya's interface, it seemed to act exactly the same as a normal map, however in a game engine this wouldn't be true. I apologize for not making that clear in the graphic I created.
All-in-all, I could go update the graphic, but that would be pointless at this point now, no?
And in further clarification, I have to say that as no-one was sure about this actually being a DISPLACEMENT map from the beginning, the whole displacement to normal map thing was not really clear as well. So simply put, mistakes happen, and that's how you learn.
Then you do know that a normalmap that is baked from a highpoly will in the best scenario be a perfect replica of the highpoly (with an exception of the silhuette) so in most cases your highpoly will be at fault.
And these normals are also baked according to the tangents of your lowpoly, and this is important to know, because if you alter these in a way such as overlaying the entire normalmap on itself, you're pushing them out of alignment, so the normalmap will no longer be correct.
Yes, but isn't it also true that certain game engines treat normal maps differently, such as the UE3 engine, where the green channel must be flipped around to be represented correctly in the engine? Is it not found by others as well that sometimes, normals don't always look like they should when ported from one program to the next?
Experienced artists provide objective educational info on the technique in their own time, which you should be happy about that, instead of busting out with "I'll never post here again" melodrama. Do you wish to understand the pros and cons of how something actually works or do you just expect everyone to love it up for you and "be on your team"?
The technique described is very old, very poor, and is now brought back as a topic by people who little technical understanding of it.
This technique is old. Thank you! I did not know that and we can all move on, I suppose. I am a learner still, and yes, I wish to learn the pros and cons. I didn't mean to give the impression of a "Melodramatic" person, and I think that may just be your point of view. It was me making a point that some of the reactions here maybe could of been a bit different, but whatever, it's a forum, that's the way it goes on the internet, I understand that, nevermind it then. It only seems to be that in learning the facts about this thing, people just can't seem to keep their wits amongst themselves and just lay the facts out, rather than saying "someones bad at this or that."
Yes, but isn't it also true that certain game engines treat normal maps differently, such as the UE3 engine, where the green channel must be flipped around to be represented correctly in the engine? Is it not found by others as well that sometimes, normals don't always look like they should when ported from one program to the next?
Flipping channels is just about inverted values, but they'll work the same way, the other normal issues have been discusses alot in another big thread here, but generally they should all follow the same rules.
Why I wanted you to post some updated images with the highpoly source was mostly because it would help us see if you did something wrong with your normalmap bake, which might very well be the case since the quality seems to have dropped on that one.
So basicly, helping you get that normalmap to work as it should, which seems to be a bit wonky in zbrush.
Flipping channels is just about inverted values, but they'll work the same way, the other normal issues have been discusses alot in another big thread here, but generally they should all follow the same rules.
Why I wanted you to post some updated images with the highpoly source was mostly because it would help us see if you did something wrong with your normalmap bake, which might very well be the case since the quality seems to have dropped on that one.
So basicly, helping you get that normalmap to work as it should, which seems to be a bit wonky in zbrush.
I will definitely get to doing that. I do think that would benefit me a bit in learning what went weird somewhere. I'll try it out tomorrow and see what's up again, too late atm lol..
step 6 says: reder disp map - what do you think this is about? its a displacement map that you enhance in ps and downgrade to a bumpmap, nothing more nothing less
in the "old days" zbrush only was able to generate displacements and such from a basemesh that gets subdivided, not like its usual, bake the data from 1 highres mesh to an entirely different lowres mesh. Basicly what you do is giving it the difference between the high and the low with a morphtarget that deforms the highres mesh.
about the greenchannel it depends on where you bake and what you options are, in most cases you are able to just switcht the YAxis to -Y and you don't need to invert green then, you just need to know where your normalmap will go to, when you render in max default settings apply for unreal, for marmoset you would need to switch, if you render in xnormal you'll have to switch the YAxis accordingly to use it in unreal but it will be right for marmoset.
Guess I'll make a last post here to back up eld. He's right. Notice he doesn't say "experimentation", he says "blind experimentation".
When I was a kid I tried to make a robot out of some wooden boards, nails and a bunch of wires. My theory was that if I tried hard enough and experimented with all possible combinations, eventually I would end up with a living talking robot through sheer force of will. That's blind experimentation.
Proper experimentation would be when you have sufficient knowledge of a subject and the experimentation part will push boundaries. The thing is, normal maps are more or less perfect, and so are displacement maps. There's nothing "more" you can do with them general-purpose wise. No matter how much experimentation you do, this will remain a fact - a fact perfectly understood by people with sufficient technical knowledge.
Tristamus, I think the best thing for you to do personally is learn how to use classic normal and displacement maps correctly. Then if you want to experiment after that, go ahead, but at least master the basics first. Luckily this forum is the perfect place to learn, so you're already on the right path there.
Alright, that all sounds good to me. I definitely have to defend myself though and say I do know how to use and make normal maps. The example graphic I made was done slightly in haste and was really only to show the raw idea of what I believed to be something that was going on with the very fine details, so by now, the graphic is actually pretty damn terrible, and I'll agree with that. Displacements, I don't really ever use so much, and that's where all my confusion comes from really, with this old way of making displacement maps being a surprise to me!
Thanks for being straight with me Perna. So, the final word I guess is that this shouldn't be used at all for anything? Or if anything, only a bump map? Only outside of game engines?
Replies
thanks to both Rastaman and tristamus
and if thats what zbrush produces as a normalmap, somethig is definitely off, did you try xnormal before? good luck with using this map from now on, might work for your portfolio but will kick you in the ass once you are in production
As far as I can see the only difference is the resolution and normalmaps are always more lowres then the actual image size
edit: and your 2nd normalmap looks really bad, can you post the maps?
example:
bump map would be good at showing the subtle pores on someone's skin, but you'd want a normalmap to help with the curvature of the skin.
the gradient/ height difference is generated between the pixels on a bump map so this can give the impression of higher res/ sharper edge lines....depends on the filtering employed i think...
and how close up you get to the map...IE a high res bump at a distance looks great but up close a low res wont give as good a result as a normal map
That is, considering the fact that the end result texels will have a normal direction, which is either generated from the bumpmap, or just accessed directly from a normalmap.
The interpolation between the two pixels are the same as interpolating between two diffuse pixels.
U can change the pixel blending to an extend where you end up with an more pixelated result. But you get sharper lines making you think the texture is more highres. Normalmaps are usually blend "physically correct" what is good but you need more then 9 pixel to define a sharp bump (e.g. a screw or something)
So when I'm talking about resolution I'm not talking about technical stuff here in contrary to what you are referring to with normal-per-pixel-resolution per.
Yes, I'd also say majority do that conversion, just like bumpmap NV photoshop filter or other stuff. You need a normal in the end, and sampling neighbors to generate it from height differences is something you don't want to do at runtime, you just want to sample the texture once and get the normal (or gradient) directly.
in the really old days, like when Per was in his fortys, "bump mapping" was "emboss bump mapping", as there was no dot-product (normal/direction angle) instuction in hardware yet. http://www.tweak3d.net/articles/bumpmapping/ (that article is from 10 years ago hehe, so it basically ended with an outlook on per-pixel lighting hehe)
+ it's simpler to sharpen the black n white bump map compared to the nm.
And this is totally on topic bc I guess that's what's going on
This is about the feeling of higher resolution not technical a higher resolution
But I have to admit that my first sentence was not clear enough defined to make it per proof.
Yes, I'm aware these maps are absolutely terrible lol, I made them very quickly without any intentions of making them look excellent. I just want the concept of whats going on here to be gotten acrossed.
Yes, I think xNormal is amazing
I actually took this map and converted it with xNormals Photoshop Filter tools, and laid it on top of a normal map from the same geometry of which I created in ZBrush, and it really did give me some amazing results.
OF all things, I'd say that this map could be purposed as a GUIDELINE to how good a NORMAL could be if made correctly for that specific object. I used it to really tweak my normals to look like the exact same thing / better as those "Special" versions, but instead as a beautifully working Normal map.
I suggest you guys don't write it off instantly as useless and play around with it. No harm can come of it really.
If you want more sharpness in your result, you have a few choices
1. Make your model sharper
2. Increase your texture resolution
3. Sharpen your normals in photoshop. Contrary to popular belief this will not explode your computer.
This whole thing is trying to throw some random shit together to make up for the fact that the artist isn't capable of getting a good result, as very evident by the example images. This is either an artistic or a technical problem, and not something that can be fixed by throwing random shit on your textures and hoping for the best.
A normal map bake is the direct result of result of the content you use to create it, the most straight forward way to create a good normal map is to create good content. Tweaking your normalmap in photoshop without really understanding how a normal map works is definately not recommended, as generally your normal map is baked to match your lowpoly's mesh normals, and as soon as you start doing crazy stuff to it, you will introduce all sorts of artifacts and errors.
Here is some reference to Dithering. If one want to use the special super duper map he should use it, but imo there are faster ways of doing it.
this too, I don't see what's new here, or better...it's a bumpmap.
please don't call that a bullshitmap though it's a bumpmap, period...marketing is everywhere these days.
waiting for magicmap though
Arguing about if it's useful it's a nonsense, play with it and find its utility. I can say it has its potential.
thanks for the tip!
QFT
in my earlier post i was talking about oldskool emboss bump mapping...and the one advantage i remember them having....
curious if any engines that used detail grey scale "bump" maps ( i remember a few being talked about a few years ago but had no experience with them) are really that if as you say they converted to normals that would seem expensive...
interesting but slightly pointless conversation as its application in realtime is mostly historic
Yeah, Doom3 is a good example, you can input a grayscale bump, but it doesnt actually use this to render, it converts it to normals and combines the two(bump+generated) on load(or compiles it into a compressed texture or something once at re-uses that).
I guess bc its funky stupid putting things on and of and doing this and that and eating bats and coocing snakes and if you HAVE the spirit you will be able to procue the special special map
it's not the first time we see something like this :poly142:
And I for my part love it
Hey guys,
I thought it was something interesting that could be played around with, that's all. Many of you are acting like I'm creating a new religion or something...chill the f**k out for a second and just look at it for what it is, maybe? I'm "Marketing" this, you say? Okay, sure thing, business man.
Like I said, use it at your own discretion. I'm not trying to prove anyone better or worse or right or wrong here (of which many of you seem to love to do in this thread).
As an ARTIST, you guys should be excited and explorative when something like this comes along, and not cynical and close-minded. Maybe some of you still have something yet to learn that goes beyond the technicalities....actually, that's quite obvious the way some of you are acting.
But anyways, back on topic, it's interesting that this kind of map can be produced out of Zbrush. Is there anyone who's played around with it at all?
Did you read the thread I linked to, at all?
RASTAMAN found this stuff, and he contacted Pixo and that's what he said they replied with.
But when i render the normal map directly from the high poly model. Shouldn't i get exactly the same amount of detail like with this special map / displacement map?
I think people are more willing to listen to someone who actually knows how things works, and I mean ACTUALLY works, like normalmaps, bumpmaps, displacements etc, but yours has been mostly guessing in the name of "exploration". You you clearly noted that the specia.. bumpmap looks way better than a normalmap when in reality you really should get better at baking and handling normalmaps, your second example even shows that your examples aren't valid anymore, since the normalmap is all stretched all over it.
But sure, experimenting is fun and all, but in the end, without raw knowledge, you'll just end up doing things you dont understand, you'll just end up misleading other people who also don't know how things work. Those mislead people just end up here and cause a headache for everyone.
Technicalities are important since computers will always follow rules, it's their very nature, you can't just randomly break them.
I shall show you this entertaining thread that popped up on gamedev one day:
http://www.gamedev.net/community/forums/topic.asp?topic_id=11259
If the maps are the same size they should represent the same amount of texels on the model surface, so they should have equal fidelity, but the normalmap baked from a highpoly is exactly that, an exact representation of how the highpoly looks, with that amount of texture fidelity.
You only need a pixel from a heightmap to create some normalmap differences, in the end it's all about how the heightmap is chosen to be intepreted.
no i didn't read the extra thread but reading your instuctions tells me its nothing new its just the way you did displacements and stuff in zbrush before they introduced their new tool for map creation - i don't even remember its name anymore because it produced bad flat normals when i last used it.
I definitely am a learning artist, like many here. Yes, I said the normal map, without any tweaks, looked less good than the special map, without any tweaks. Without any tweaks. It was a raw comparison, visually. I also stated the examples maps I made were done quickly and were crap....I was just getting the concept across about what I thought was going on. And, who said I was talking about using this in just video-games or in production animations etc? If it looks good in a still shot / render, it looks good in a still shot /render, no?
Yeah, computers follow rules, and then you find glitches like this. What's your point, man?
Why am I arguing? I came to show that this exists. If you don't like my opinions on it, that doesn't really matter. I'm willing to learn more, but telling ME personally that "I don't really understand technicalities" and basically telling me no-one should listen to me is absolutely besides the point. When did I ever say "YOU MUST ALWAYS FROM NOW ON USE THIS MAP. USE IT NOW. I COMMAND IT" lol....
So, what I'm saying, is stop the *I'm better than you* routine and just get on with our lives here lol...I simply am having to defend myself from...I don't know what, exactly? Smarty pants people who are trying to talk down to me? I'm not sure really...
I simply thought this would be something interesting the community could play around with?
God forbid I ever show anything here again...
There's also the people who would overlay a normalmap unto itself since they heard it was a good trick to enhance the looks of it!, that kind of blind experimentation mentality isn't pretty, since it never gets anywhere.
I wasn't refering to the quality or anything about your examples, I was refering mostly to the fact that the pillar clearly isn't mapped the same on both examples, and it would help if you showed the actual highpoly you baked from.
We're not doing the "better than you" routine, but you do have to accept, this is polycount, good things comes out from listening to the people who have experience with these kind of things on these boards.
If you go through the trouble of modeling high res geometry you better be baking it directly to a normal map. Rendered normal maps define the exact difference between the low res geometry and the high res geometry which is needed for correct rendering. If you were to try to render a displacement map from these 2 objects and then try to convert it to a normal map you will end up with horribly inaccurate results, which will result in lighting errors/seams/artifacts in your renders.
I completely agree. But, blind experimentation?
It has also been made very apparent to me that mixing images and maps in photoshop is NOT something frowned upon. Why would you say...
"There's also the people who would overlay a normalmap unto itself since they heard it was a good trick to enhance the looks of it!, that kind of blind experimentation mentality isn't pretty, since it never gets anywhere."
I don't agree with you there...
I know MANY professionals currently at game studios, of whom used to be professors of mine, who would do that kind of thing regularly and suggested that it was something that you needed to TWEAK / EXPERIMENT with, and that great results could come of it. In fact, tweaking is experimentation in itself, is it not?
My whole point is leading to the fact that experimentation is good, and in me bringing this over to polycount I hoped to see some people experiment with it to see what cool things could be done. By saying something along the lines of "Blind experimentation doesn't get anywhere," not only are you putting aside the fact that this kind of thing could possibly be useful somewhere, you are being oblivious to the fact that experimentation IS a professional way to go about things, and by those means, it should be encouraged.
I will also say that I would love to listen to what professionals have to say here on these forums, pertaining to this kind of thing. Perhaps someone who does not have to talk down to a person whilst teaching them a thing or two?
I completely agree there, concerning normal maps and such. However, is this map really a displacement map? I don't work with disp maps too often, because they dont work in real-time. They do look different do they not? Or am I wrong? Also, it doesn't HAVE to be converted to a normal map. Could it work for anything else? That's what I'm also interested in...
People have used various greyscale maps (curvature, displacement maps, dirt...) to enhance their textures, is nothing new and common practice.
what you suggested (the really bad normalmap bake picture in your first post) is that normalmaps are inferior to this "special map" (displacement -> normalmapfilter) workflow. And that is very very bad, as stated by numerous people here.
And because that is your main point (first post) and the main gist of this "marketing" like picture (really crappy normalmap bake vs what a normalmap normally looks like "special map"). That is what upsets people. Not the fact that you experiment with greyscale or whatnot maps.
Then you do know that a normalmap that is baked from a highpoly will in the best scenario be a perfect replica of the highpoly (with an exception of the silhuette) so in most cases your highpoly will be at fault.
And these normals are also baked according to the tangents of your lowpoly, and this is important to know, because if you alter these in a way such as overlaying the entire normalmap on itself, you're pushing them out of alignment, so the normalmap will no longer be correct.
Let me say now, that when I stated that I did not know what the "Special" map was all about, it was still just starting to get around to quite a few people at ZBrushCentral and myself. It's definitely old now in comparison to the new knowledge gained from this place. The fact is, in Maya's interface, it seemed to act exactly the same as a normal map, however in a game engine this wouldn't be true. I apologize for not making that clear in the graphic I created.
All-in-all, I could go update the graphic, but that would be pointless at this point now, no?
And in further clarification, I have to say that as no-one was sure about this actually being a DISPLACEMENT map from the beginning, the whole displacement to normal map thing was not really clear as well. So simply put, mistakes happen, and that's how you learn.
Yes, but isn't it also true that certain game engines treat normal maps differently, such as the UE3 engine, where the green channel must be flipped around to be represented correctly in the engine? Is it not found by others as well that sometimes, normals don't always look like they should when ported from one program to the next?
This technique is old. Thank you! I did not know that and we can all move on, I suppose. I am a learner still, and yes, I wish to learn the pros and cons. I didn't mean to give the impression of a "Melodramatic" person, and I think that may just be your point of view. It was me making a point that some of the reactions here maybe could of been a bit different, but whatever, it's a forum, that's the way it goes on the internet, I understand that, nevermind it then. It only seems to be that in learning the facts about this thing, people just can't seem to keep their wits amongst themselves and just lay the facts out, rather than saying "someones bad at this or that."
Flipping channels is just about inverted values, but they'll work the same way, the other normal issues have been discusses alot in another big thread here, but generally they should all follow the same rules.
Why I wanted you to post some updated images with the highpoly source was mostly because it would help us see if you did something wrong with your normalmap bake, which might very well be the case since the quality seems to have dropped on that one.
So basicly, helping you get that normalmap to work as it should, which seems to be a bit wonky in zbrush.
I will definitely get to doing that. I do think that would benefit me a bit in learning what went weird somewhere. I'll try it out tomorrow and see what's up again, too late atm lol..
in the "old days" zbrush only was able to generate displacements and such from a basemesh that gets subdivided, not like its usual, bake the data from 1 highres mesh to an entirely different lowres mesh. Basicly what you do is giving it the difference between the high and the low with a morphtarget that deforms the highres mesh.
about the greenchannel it depends on where you bake and what you options are, in most cases you are able to just switcht the YAxis to -Y and you don't need to invert green then, you just need to know where your normalmap will go to, when you render in max default settings apply for unreal, for marmoset you would need to switch, if you render in xnormal you'll have to switch the YAxis accordingly to use it in unreal but it will be right for marmoset.
Alright, that all sounds good to me. I definitely have to defend myself though and say I do know how to use and make normal maps. The example graphic I made was done slightly in haste and was really only to show the raw idea of what I believed to be something that was going on with the very fine details, so by now, the graphic is actually pretty damn terrible, and I'll agree with that. Displacements, I don't really ever use so much, and that's where all my confusion comes from really, with this old way of making displacement maps being a surprise to me!
Thanks for being straight with me Perna. So, the final word I guess is that this shouldn't be used at all for anything? Or if anything, only a bump map? Only outside of game engines?