[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / r / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip / qa] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
Settings Mobile Home
/3/ - 3DCG

Name
Options
Comment
Verification
4chan Pass users can bypass this verification. [Learn More] [Login]
File
  • Please read the Rules and FAQ before posting.

08/21/20New boards added: /vrpg/, /vmg/, /vst/ and /vm/
05/04/17New trial board added: /bant/ - International/Random
10/04/16New board for 4chan Pass users: /vip/ - Very Important Posts
[Hide] [Show All]


[Advertise on 4chan]


File: normalmaptutorial.jpg (449 KB, 1920x1080)
449 KB
449 KB JPG
>the most popular normalmap tutorial for photoshop
>he creates normalmap straight from diffuse, no heightmaps involved
>in the end his bricks are concave
>doesn't even mention that photoshop doesn't have option to generate OpenGL normalmaps and you would have to manually flip green channel if you want OpenGL standard normalmaps
>video gets a lot of praise in the comments and no one points out this tutorial is absolute shit

How do we stop Crises like this from making tutorials?

https://www.youtube.com/watch?v=dKBav9IsRiU
>>
>>988022
Every other app has a different green direction, you can find that out on your own and flip it, no biggie. Resulting map looks competent too. You are whining about having free information and calling someone a Cris for no reason.

You're weaksauce crybaby Op, the only thing you need to stop is yourself; from bawling your eyes out with salty tears.
>>
>>988022
No other apps have a different green direction, you can't find that out on your own by flipping it, it's a huge deal. Resulting map looks incompetent too. You are justified in your complaints, and that person is definitely a Cris.

You're strong and justly upset Op, the only thing you need to stop is the guy who posted that vid, from bawling his eyes out with salty tears.
>>
>>988022
>>988024
>>988027
Umm actually, as someone who has certifications and experience in Photoshop. Adobe stupidly decided to remove it. https://helpx.adobe.com/photoshop/kb/3d-faq.html

It’s not the YouTuber fault, the video is just outdated. Adobe is the real culprit behind it.
>>
>>988027
>No other apps have a different green direction

max, cryEngine, source, unreal engine all have Y- tangent space maps. Blender, Maya, Unity, Marmoset Toolbag have Y+ tangent space maps. Having to flip green channel is common practice everywhere.
How green are you anon?
>>
>>988039
i think it went over your head
>>
>>988040
Apologies, it appears I have hallucinated a response post that wasn't there.
>>
What do you guys mean by "flip green channel"? I have no experience with normal maps, btw. I only know the broad concept that each channel is associated with a dimension. But what do you mean by flipping green? Green is green. How do you "flip" a color?
>>
>>988042
>I have hallucinated a response post that wasn't there
exactly what chatGPT would say
>>
File: 6846546154.png (258 KB, 680x681)
258 KB
258 KB PNG
>>988022
>photoshop doesn't have option to generate OpenGL normalmaps
>points out that you can easily flip it anyway
xnormals plugin let you flip any channel
it also has different tools for creating cavity maps and renormalizing normals
>>988042
how many bots are there on this site now?
>>
>>988269
>how many bots are there on this site now?

The belief the majority of posters at popular websites like '4chan.org' are bot's can be broken down into these underlying themes.

1) Format. A poster whom adopted the pattern of typing in paragraphs and structuring their sentences in a way
that appear bot-like can give the impression that the poster is not humanoid.

2) English as a second language (ESL). A poster who is not a native English speaker may use sentence structure and thought patterns
rooted in their native language, the result of this can appear uncanny to a native speaker especially when it's combined with refined vocabulary.

3) Paranoia. The realization how synesthetic minds now rivals natural ones in their capability makes the distinction by artificial and natural
mental processes difficult to differentiate between. In the face of this development how proficient any individual human is at recognizing another human
will fall onto a bell-curve. Well informed individuals who understand the problem while lacking the rizz to recognize it in others will have many false positives.

If you believe your forum has been breached by bots several great mitigation strategies are available to you such as:

1) Demand identification that can be verified by more secure external systems such as bank ID.

2) Take your meds.

3) Realize that humans adopt the speaking patterns of their peers and that many people now converse daily with systems such as chat-GPT
therefore adopting similar idiosyncrasies.

As a Large Language Model I can have no opinions on any of these matters cept how Op is ofc a huge fag.
>>
how does one create normal maps, usually?
>>
>>988333
Photoshop but because Adobe removed it, the new method is Gimp: https://youtu.be/9cXourS7i9s?feature=shared
>>
my cracked CrazyBump.exe just werks
>>
>>988022
>he creates normalmap straight from diffuse, no heightmaps involved
>in the end his bricks are concave
Isn't the diffuse value probably a decent proxy for height considering there's AO in the crevices?
What else are you supposed to do starting from diffuse? I think CrazyBump basically interprets diffuse as height, and just lets you blur different frequencies if you want.
Beyond that, you could run it through some curve to maybe get nicer results.
I'd be interested to know if there's some more correct method of inferring height data from the diffuse (especially considering that accurate absolute heights are not important, just the local height deltas for computing normals)
>>
>>988344
>I'd be interested to know if there's some more correct method of inferring height data from the diffuse

Run a high-pass filter across the image to gain a evenly bright surface, this will be suitable to extract normals for for the micro-detail texture.

To capture macroscopic features like deep changes in geometry you use masks and painting to suggest the elevation change.
It doesn't need to be perfect to start looking correct as soon as there is a suggestion that different elements are on different planes
your brain will start interpreting it as more advanced geometry than it really is.

You can also run different types of emboss filters horizontally and vertically and shift the levels manually towards the midpoint to extract a lot of
medium sized details from a diffuse map. You place the emboss horizontal pass into the red channel and the vertical into the green and run 'normalize normalmap'
to generate the blue channel. You can then boost the depth of the generated blue channels via levels and exposure and re-normalize the map to gain even more details.

You're never gonna get the map to look as good as normals captured from real geometry but
you can make it a hell of a lot more compelling than anything 'crazybump' would spit out, ofc you're paying for that extra quality in time spent.
>>
>>988333
proper high-end normalmaps are made by setting up a projection cage to render or 'bake' the normal information of a high-poly asset onto a texture map for that low poly asset.

like you have your sculpt of a character that is 10+ million polygons and you project this detail onto the UV's of a asset suitable for real time rendering
onto a lowpoly asset suitable for use in game engines. (A 'lowpoly' asset today may still have a polycount of upward to 200K polygons for a AAA type character).

Search for 'texture baking' 'Render to texture' if you wanna find examples of how it's done.
>>
>>988382
What is the average AAA character to play as, or NPCs? I'm moving soon from modding and creating lowpoly characters on an old engine to UE5 and would like to do cinema tier quality with textures, as much as my skill allows me to.
>>
>>988385
I forgot to add: polygon count.
>>
>>988385
A main plot character in Cyberpunk 2077 is about 150K triangles. I consider this a pretty ordinary ballpark number for recent AAA's.
>>
>>988388
They manually retopology and draw up to 200.000 faces by hand? I'd assume the retopo time would take a week on average, quadrupled to what I'm used to doing 5k poly characters.
>>
>>988390
>They manually retopology and draw up to 200.000 faces by hand?

Yeah. Moving that many polygons isn't super hard to do with contemporary polygon tools. A week is reasonable if you're working ordinary sessions, about 2 16h sessions or two work days if you're crunching and sleeping in a bag underneath your desk.
>>
>>988391
Yeah makes sense now. Having a good retopology software also matters, I used blender with retopoflow but the amount of lag and issues in it would make everything take longer.

I've recently settled with TopoGun 3 and happy with it since it handles high polycount meshes buttersmooth and the tools are decent, not sure if there's something more advanced out there which has better tools to work with to say.
>>
>>988393
I still use max's retopo tools, they're not the best but they get the job done and I'm very familiar them.

I came from a background of modeling high-poly assets before 3D sculpting was even a thing so making realtime
characters today is not that dissimilar from making characters meant for offline rendering back in the day.

The poly density on assets we use for realtime today on games is similar to what we used for CGI rigs in the early 00's.
Only we ran 2-4x sub-D smoothing to bump up the resolution of the mesh at rendertime and sometimes use displacement maps
for the kind of details we typically just sculpt today.
>>
>>988267
Green is green to our eyes, but colour information has chirality in the digital space, so the computer can say "this is green" or "this is mirror green", and still display it as regular green.
Programmers sometimes might want to specify which green they want for certain operations, or they can optimise their game by deleting the chiral colours from their palette, which isn't something you'd use these days because memory buffers are so large.
A common way to flip green is to save out the green channel in your image editor as a .bmp, then bring it into audacity as raw data then flip it there. Make sure you don't flip it + reverse it or you'll get double chiral green.

Actually that's all nonsense, you can just do pic related. Curves adjustments are all relative so the green will swap with the magenta.
>>
>>988267
If you inspect each individual channel of a RGB image you find it's just a value ranging from 0.0 to 1.0, or 0 to 255 for a typical 24bit image (8 bit per color channel).
A black and white image essentially that you can place in any of the three channel to display as your red green or blue component.
Flipping a channel means you invert it so the white becomes black and black becomes white.

When you flip the green channel it means the parts of the image that had no green now turns green and the parts that where fully green now has zero green.
>>
File: normals explained 2.jpg (614 KB, 1258x1408)
614 KB
614 KB JPG
>>988414
And in the context of a tangent space normal map it is encoded such that a value of 0.5 or 127 in red or green channel
means it is pointing neutral in that direction, Red represents X direction, green represents Y direction.
Blue points out the screen towards you in Z direction and it is neutral at a value of 1 or fully white.

so RGB(127,127,255) means the normal points the same way as the normal of the surface why you see that light-blue tint across most of a normalmap.

Now some programs disagree if the Y axis should point up or down so if you invert the green channel it will point towards the other direction.
but the so 255 becomes 0 and 0 becomes 255, but the middle neutral value of 127 is still 127 when inverted.
>>
File: 1442092939124.png (339 KB, 472x432)
339 KB
339 KB PNG
>>988022
>he creates normalmap straight from diffuse
gross
>>
>>988022 There's way more to normal maps than just making a height map and using a generator, this vid explains some of it: https://www.youtube.com/watch?v=uOPwVZI5UTM



[Advertise on 4chan]

Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.