So not only do your videos show off making stuff look good in video games, but it also shows off making the illustrations in the videos themselves look good. The animations illustrating the concepts you're describing starting at around the 1:18 mark are particularly exquisite. I'm here to tell you that the time, polish, and effort you put into these videos have not gone unnoticed! Well done.
I've just recently discovered your channel. I've put shaders off till the end. But I think every math professor would be so jealous of your ability to explain these abstract concepts. Thank you
Maaaan, thank you for *finally* describing this in a way I can understand. I could never understand why a "flat" normal map was 0,0,1, but makes total sense now when you say we're multiplying the normal map values with the tangent/bitangent/normal.
For me the most important take-away from this is the basic explanation of why normal maps are so important and what they do. The light bulb just clicked on, "that explains a lot of things".
the best part of this is . . . most users of 3D creation software don't have to understand one bit of what you said . . . but we are eternally grateful that people like you do. thank you.
Your videos are a treat to be honest, and your case study videos are eye opening to all the possibilities that can be done through shaders. How does one acquire a deep knowledge and understanding of shaders as yourself? How did you learn this stuff? There's no other channel or resource that is half as good.. and the blogs tend to be introductory, and no one tackles 2D. I hope you could shed some light.
This shit literally blew my mind... this legit made me realise that the games I play use flattish surfaces and they just have fucking wild normal maps so it lights differently... BLEW MY MIND
Thank you for starting this series, I have been looking into shader writing for a while now, but it's so hard to find a collection of useful info anywhere, so thanks a ton!
I'm not going to complain about this video because I didn't understand it. The lack of knowledge is mine to own. I felt if I watched it with the pause button looking up what I didn't get and replaying the bits I'd almost got then I'd really understand it. I've watched some terrible tutorials that really are just someone showing off. This is not one of them. Thanks. By the way, I do understand way more than I did before watching it. Big thanks.
Speaking of normals and lighting. I think you should enlighten developers about usage of MatCap shaders and their advantage in mobile development. I've found them extremely useful.
This was excellent. I'm doing a adv dip including OpenGl, and this is something we've recently covered. Except yours was a much better resource than theirs. ^.^
This was extremely informative, and i'm not even using Unity, so though I filtered and saved for later, a lot of what you said, this helps me finally GET wtf normals are, why they're the colors they are (I was like, is there some standardized gamut everyone is using?) and HOW they work. even though directionality can change by app e.g. in Keyshot vs Blendr, there's plenty of transforms and toggles to fix em. my main confusion was why they were even necessary when you have a bump map, geometry, and in some cases a displacement map. FINALLY i feel a little less dumb today. :) thank you!
you explained dot product and I was like "cool I wondered what exaclty dot product was doing" then you said vector to a lightsource from a vertex and it was like the techno wizzardry that is how 3d models are lit became clear.
The tangent was a little unclear tho. Am i right if i simply say that the 3 values ib rgb are storing the vector normal to the surface. It may be relative to somthing but i cant tell. I need to give a little bit more time
I came here to finally understand what's a normal map... and not only I didn't understand a word... but now I am even more afraid to dig in. Yes, I had problems at school and my maths level are a that of a child, but there are a lot of people like me... we also have the right to know!! Hahahaha...
I'm a bit late, but your definition of the dot product is only true for unit vectors. The dot product of two like vectors is the square of their shared length. That being said, normals will of course be unit vectors, but I mention this for the sake of accuracy.
No plans for parallax mapping in the future, but if you have GDC vault access (or you can get access through a friend/work/whatever), part of the talk "Shaders 102" covered a similar technique in Unreal. The talk as a whole is worth checking out. And stencil buffer ahhh.... I used it in my See through effects video but it's not the best use case for stencil stuff. Compelling reasons to use stencil buffer are pretty few and far between as most ways to use it can be achieved with the depth buffer already. I'll try and think of a cool way to use it for a future video hopefully!
I know these might be really easy but maybe could be a quick video that gets good traffic (because of the game). Maybe you could do a quick case study on the enchanted items icons in and the end portal in Minecraft. And whatever other cool shader stuff is in that game...Can't think of any off the top of my head. Love the depth of knowledge you show on your videos ...i only understand some of it but it reveals there is a lot I didn't know i didn't know. Maybe these (what i presume to be) simple shaders or tricks would be something a broad lower level (relatively speaking to yourself) audience might want to implement and can be knocked out relatively quickly. Also maybe you could consider doing a video showing us the best way to interact with shaders via script such as turn them on/off like in the end of the pokemon dive or modify them with gui controls or through user selection, like you did in your web demo of the Spelunky shader..A unity talk on mobile optimization has me paranoid about mobile optimization..if you watch it you will understand why ahah but he talks specifically about shaders at 34:35. czcams.com/video/j4YAY36xjwE/video.html. Maybe you could talk some about shaders and running them on mobile. Just some thoughts. Probably the most technical in depth unity information with professional content on youtube I've found . Great work hope to see more soon.
"I'll get into art as I recover burnout after studying computer science in university. I'm sure everything will be just very calm and no math involved. :)" (jokes aside really good video but lol)
How would one create normal data if they don't have the original objects but they do have a texture? Note: This is a technical question, not a "use application " question.
Awesome video. learned a lot about normal mapping because you just showed it how it is. Best way I learn things :D What is that thing with the top hat? Looks like a state of the US?
I Have an Idea what you could analyse next. Prey's Looking Glass technology... it's basically the same as Portals in Portal but still very interesting and a shader break down would be great
Am I just crazy or are you using Donkey Kong Country Soundtrack as background music (5:41 sounds like stickerbrush symphony)? Edit: Lol just saw the Video notes and I'm right :-)
I would love if you went back to the exercises you had in shaders 101, it helps reinforce what you are saying so I am not just listening along. Keep up the awesome videos though!
at 4:50 you show the matrix multiplication. shouldn't you do mul((float3x3)unity_ObjectToWorld, v.normal) instead? You use the inverse matrix which also happens to be the transpose on the right side, matrix multiplications are not commutative so while in this particular case it works and it does transforms from object to world space it might not always be the case. If I'm not mistaken unity objects scale does some weird stuff if you get it wrong, besides, its the correct way to read it.
Yeah I probably shouldn't have moved past that so fast. What you've written here will work but only for uniform scaled transforms. Right side by the inverse will work properly with scaling. It's been a while since I've looked at the proof for that, I'll try and find a good source and make an annotation. Thanks!
Hey! I know this is a bit off topic to your current video, but do you know of any smart way of masking off the world to only render objects that are inside a sphere? I've tried using the stencil buffer but the problem is that if you angle the camera you will continue to see the object despite of it not being physically inside the sphere... I need to achieve a type of snow globe effect, like a little world inside a sphere☺️ Thanks for the awesome videos! Keep them coming👌🏼
Not entirely sure what you're trying to achieve, but a sphere can be defined by just a point and a radius. So if you're writing/editing the shader of the objects that are within the globe, you could pass through world space positions from vertex shader and use clip/discard with the globe's point and radius.
Question. Question. Question. At 4:50, shouldn't it be the ObjectToWorld matrix??? And also, shouldn't the matrix be the first parameter? Is the opposite order of the parameters what causes the matrix to be the inverse?? So many questions.....
Very good questions! I don't know why I skipped over this. We are transforming a normal, which is slightly different than just transforming a direction (that's what the math you're suggesting performs). You need to multiply a normal by the Inverse-Transpose of your model matrix to properly account for non-uniform scaling. If you look up why you have to do that you'll come across a stack overflow page with several proofs and different explanations, but the goal is to not squash/stretch your normals if you're scaling non-uniformly. Then a little trick that Unity actually uses in their cg include files, is to perform the multiplication on the right side. This is the equivalent to multiplying by the transpose from the left, except that mathematically you'd be multiplying a 3x1 matrix instead of a 1x3. But shader compilers don't care, they let you use a float3 as a column or a row contextually. So the inverse of ObjectToWorld is WorldToObject, and multiplying from the other side accounts for the transpose, which all in all gives us the effect of multiplying by the inverse-transpose. If you can assume uniform scaling, you can just use ObjectToWorld from the left-hand side as you have suggested and your normals will look just fine :)
Thanks for your response. This actually motivated me to spend 3 hours reading up about the whole matrices things. Thanks for being informative as always. Also I just made a video showing off some of the shaders I wrote. You can check it out on my channel if you want. No pressure tho.
Makin' Stuff Look Good I'm glad you like it. It looks pretty cool. It supports tessellation so you can effectively swap out the model by baking 2 different models on the same base mesh. I'll probably make a video highlighting each of them. Again thank you for your feedback :)
Could you make a video on Terrain shaders? One with triplanar for dealing with texture stretching and UV resizing by distance to reduce patterns?( that would be amazing, lol). Anyway, there's hardly any info on terrain shaders for unity in the internet. Most free terrain shaders out there are 4 to 5 years old and don't work anymore. So it doesn't matter if you just want to use them or learn from them, because they just don't work anymore. So I was thinking it would be of great use for many people.
Just my opinion but it would have been better if you used triangle for 3 point plane, as triangle is a figure which can be drawn only two ways given 3 points (the only difference is which way normal is pointing), is why it's used everywhere instead of something like a quad or hexagon, as other figures don't have the same property. Also it's not hard to calculate normals yourself, even though it's useless if you're using an engine, since it will always have a way because calculation is simple and works for any valid mesh.
There are only three points, but 4 are visualized. The fourth is actually point A reflected across BC, that's why there's no control handle for it and it moves on it's own. To visualize a plane, what I really wanted was one that stretched out to infinity like a plane actually should. But that visualization didn't read well so I changed it to a quad.
Gamma correction has barely nothing to do with human vision... That's a misconception / myth, and it's making me angry whenever I see it propagated further. Our eyes, just like a camera's sensors, record values in linear space. We see linear values. We need linear values. We never see gamma values. The only reason for gamma correction is backwards compatibility with old CRT technology. Back then, the monitor tech was so bad that it was displaying darker values than those expected and supplied to the GPU (coincidentally, in a non-linear fashion). Which meant that a value of 120 was displayed as less than 100 (not accurate, but just for example). People figured out that by supplying higher values than originally intended, they could get the monitor to display the required brightness. To do that, they started to gamma encode values to higher values (which again, works in a non-linear fashion). By supplying these gamma encoded values, the monitor was (through the nature of it's cathode ray tube) capable of decoding those values to the required lowered ones (keep in mind that this decoding was an abstract, conceptual decoding. There were no actual mathematics performed by the monitor) After the dawn of CRT technology and the birth of LCDs, monitors were finally able to correlate the GPU supplied data with on-screen brightness (which meant that 120 was 120 on the screen). The biggest issue, however, was that the ENTIRE INTERNET was filled with gamma encoded images, and backwards compatibility was required (there was no way to purge all the servers and change the standard to no gamma correction without making all images up to that point obsolete). "Why is that?" - you might (hopefully) ask. Well, if you try to display a gamma corrected (encoded) image on an LCD with gamma correction features disabled (gamma = 1.0), you'd get a very bright and washed up image (remember, gamma encoded / corrected values are higher than those intended). LCD manufacturers figured that the only way to keep going forward was to cancel gamma correction from images at software + hardware level. The LCD has a LUT (LookUp Table) chip which is responsible for doing gamma expansion / decoding, effectively canceling gamma correction and displaying values as originally intended (basically doing, through mathematical means, the same thing that a cathode tube was doing naturally). To enable this decoding process, you need to set your monitor to a gamma of 2.2 (the number was chosen as an average between 2.0 - 2.4, during the old CRT days, because it was giving the best results for a wide range of monitors), because almost every artist is encoding using that value. Of course, panel quality is very important here, so a value of 2.2 might not provide the exact experience intended by the artist (the LCD panel might be bad at reproducing colors / have bad background lighting). Basically, the entire process is like this: an image is created (values are in linear space) -> then gamma encoded (non-linear space) -> stored somewhere. Somebody comes along and downloads said image -> opens it -> (OPTIONAL STEP) if they want to alter the image, they have to do all their calculations by first decoding the image (getting it in linear space), then calculating, and after that re-encoding (back to non-linear space) -> gets sent to the GPU -> GPU sends it to the monitor -> the monitor uses its LUT chip to decode the image (linear space) -> image is displayed. Gamma correction is at this point a useless vestigial artifact of a bygone era, on account of bad early tech.
hey nice video but is it possible to explain why more about the blueish purple color being that color? eg. when we divide 255/2 = 127.5 = results in 128 for Red and the Green Channel in the mid value but for the blue when we divide 255/1 = 255 we get a (128,128,255) value. SO my question is is 127.5 = 128 or does the dividing the R and G channels should be 256/2 = 128? thanks
So, to be clear... the normal map doesn't contain raw normal vectors. It contains what percentage of the normal, tangent, and bitangent go into the final normal, and the normal and tangent will be provided in the geometry data. Is this correct?
Then I'm confused as to how moving the light to tangent space works. There, it appears the normal in the texture is the real normal. I understand that the full math behind tangent space is beyond the scope of this video, but it would have been helpful to see some sample values in the normal map, and show what the values are for the normals in world for tangent space.
Hey! You have shared some shaders in your assets - VisualObjectSpaceNormals.shader is the same as VisualWorldSpaceNormals.shader... probably you forgot to delete "UnityObjectToWorldNormal()" from your vertex-shader ;) Great video, thank you very much! I am waiting for some video about parallax mapping!
Whoops! When I was recording parts for the video I would just quickly change something to get the shot I need instead of changing which materials were applied so I probably goofed. Thanks for the note, I'll fix it up later today!
If using forward rendering, why would you want to transform per fragment normals from tangent space into world space?Isn't it better to transform light direction/position from world space to tangent space instead? It would save operations in the fragment shader.
9:06 - yup I mentioned that. It's generally easier to think about and visualize the normals in world space so I explained it this way. As well, there are effects that require world normals anyway so it's good to understand how you would get them out of a normal map.
To be consistent with the rest of the video, that diagram is of a left handed coordinate, with Z point towards us and Y point up. B and C's x values are actually lower than A's. Check this: imgur.com/J73qvjn ab x ac is blue, and ac x ab is red. Sorry for the confusion!
Hey man, any idea on how Hearthstone make those green energy auras around selected cards? Im not sure if it isnt just animated texture/sprites or if theres an actual fancy effect, or maybe a combination of animation + some gloom.
How do you calculate the normal at a vertex? If a normal is usually found with a cross product, are vertex normals just averages of the normals of the faces around them?
@5:09 could you explain this bit more? How did you get that texture? I’d really like to be find a good way to take 3D animations and bake the, in 2d with normal maps for games.
I mention that around 9:00 that you can instead transform the lighting direction into tangent space and then you do a few less ALU's in your fragment code. There's an example of that in the github project. As for screen space normals, that would be the approach if you we're writing a deferred shader. If you dig into the Unity cg cinclude files they have some good examples of that!
Yeah, Unity has some support for Forward and Deferred out-of-box. In this video, we're writing a forward base pass with just one directional light. Things can of course start to get much more complicated as we support more lights/lighting features. But to keep the video from being an hour long, I tried to keep it simple :D
you're making a great difference for beginner shader programmers community... keep up
So not only do your videos show off making stuff look good in video games, but it also shows off making the illustrations in the videos themselves look good. The animations illustrating the concepts you're describing starting at around the 1:18 mark are particularly exquisite.
I'm here to tell you that the time, polish, and effort you put into these videos have not gone unnoticed! Well done.
Aaron Misner
Well said
Totally agreed. In fact im curious how you produced these. Are these Unity scenes as well?
These are high quality teaching videos.
I've just recently discovered your channel. I've put shaders off till the end. But I think every math professor would be so jealous of your ability to explain these abstract concepts. Thank you
Also, if you ever do one on linear algebra/matrix multiplication, etc, I'll be one of your biggest fans.
Maaaan, thank you for *finally* describing this in a way I can understand. I could never understand why a "flat" normal map was 0,0,1, but makes total sense now when you say we're multiplying the normal map values with the tangent/bitangent/normal.
Didn't understand 70% of what you just said... feels bad man.
feels bad too
thank god I'm not the only one
they say, "I can explain it to you but I can't understand it for you". The video content is perfectly explained
@@stevejones9044 for some one that alrady knows some data about it, there are many words that people would not know. Still is a good video.
For me the most important take-away from this is the basic explanation of why normal maps are so important and what they do. The light bulb just clicked on, "that explains a lot of things".
the best part of this is . . . most users of 3D creation software don't have to understand one bit of what you said . . . but we are eternally grateful that people like you do.
thank you.
awesome video dude!
the way you make your videos is perfect, don't change anything!
funny and highly educational.
Your videos are a treat to be honest, and your case study videos are eye opening to all the possibilities that can be done through shaders. How does one acquire a deep knowledge and understanding of shaders as yourself? How did you learn this stuff? There's no other channel or resource that is half as good.. and the blogs tend to be introductory, and no one tackles 2D. I hope you could shed some light.
I am doing research for my PhD and this is 10 times better than any scientific paper i have read about Mikktspace normals. Thanks a lot
finally somebody talks about how to orient the normals relative to texture coordinate
As it happens, normal maps are on my to do list for my dissertation! good timing as they're going in in a few weeks :)
Love to see new videos! Glad you still have a little free time with your new job!
This shit literally blew my mind... this legit made me realise that the games I play use flattish surfaces and they just have fucking wild normal maps so it lights differently... BLEW MY MIND
Thank you for starting this series, I have been looking into shader writing for a while now, but it's so hard to find a collection of useful info anywhere, so thanks a ton!
Just watched your entire Shaders videos, thanks! I'm still super lost, but at least you encourage me to learn more about the subject :)
I'm not going to complain about this video because I didn't understand it. The lack of knowledge is mine to own. I felt if I watched it with the pause button looking up what I didn't get and replaying the bits I'd almost got then I'd really understand it. I've watched some terrible tutorials that really are just someone showing off. This is not one of them. Thanks. By the way, I do understand way more than I did before watching it. Big thanks.
I feel like such a nerd understanding and feeling gratitude for this help
Amazing video, best tutorial I could find on here that explains normal maps
Name of the channel absolutely delivers.
Cool, dense material but cool. I'll need to watch more of your videos.
Very educational! 👏
what a brilliant explanation....well done
"Tangential explanations" - I see what you did there
Thank you for that. I've been 3D modeling for years and never really got it till now. Keep it up!
Thanks for the video man, i was always to scared to dive into shaders since i'm clueless about the fundamentals but this video helps out a great deal!
You should do a shader case study of Super Mario 3D World. This game is FULL of little details that make the game beautiful and some are... Amazing
Speaking of normals and lighting. I think you should enlighten developers about usage of MatCap shaders and their advantage in mobile development. I've found them extremely useful.
Normal map
This is your daily dose of Recommendation
Omg, this video helped so much. The tutorial wasn't 90% "do this because MAATHHHHH". Thanks! :D
Great video! Thank you. Adding the github resources is such a nice touch!
Thank You, this was very informative to me as 3d artist. Now I see what I have to look for preparing normal maps to be used in Unity.
This was excellent.
I'm doing a adv dip including OpenGl, and this is something we've recently covered. Except yours was a much better resource than theirs. ^.^
This was extremely informative, and i'm not even using Unity, so though I filtered and saved for later, a lot of what you said, this helps me finally GET wtf normals are, why they're the colors they are (I was like, is there some standardized gamut everyone is using?) and HOW they work. even though directionality can change by app e.g. in Keyshot vs Blendr, there's plenty of transforms and toggles to fix em. my main confusion was why they were even necessary when you have a bump map, geometry, and in some cases a displacement map. FINALLY i feel a little less dumb today. :) thank you!
Finally, feeling good about my year of Linear algebra
LOL HE IS ACTUALLY ALIVE
please do more! amazing work!
Awsome. This is a fantastic video, thanks a lot! Subscribed
you explained dot product and I was like "cool I wondered what exaclty dot product was doing" then you said vector to a lightsource from a vertex and it was like the techno wizzardry that is how 3d models are lit became clear.
Just realised that bramble blast from donkey Kong country 2 is playing in the background
Excellent yes I understand everything now thank you
Thank you, this was very helpful.
You are an awesome human being. Thanks for sharing knowledge! Respect.
Great stuff. Thanks for making these.
Awesome video! Would you be able to do one that shows off how Heroes of the Storm does their death & waiting-to-respawn screen effects?
My brain usage reached at 15%, while watching this.
Omg someone who explains shit. Not just saying we put this there and that there. This is how u explain things.
The tangent was a little unclear tho. Am i right if i simply say that the 3 values ib rgb are storing the vector normal to the surface. It may be relative to somthing but i cant tell. I need to give a little bit more time
great video !! Please keep uploading new videos
I like the intro
I came here to finally understand what's a normal map... and not only I didn't understand a word... but now I am even more afraid to dig in. Yes, I had problems at school and my maths level are a that of a child, but there are a lot of people like me... we also have the right to know!! Hahahaha...
such a good explanation!!
love some more GLSL scripting tutes
[Previous knowledge of texturing terms necessary to understand anything]
Finally understand the dot product.
Wonderfully explained them on math. Thank you for this usefull clip.
I'm a bit late, but your definition of the dot product is only true for unit vectors. The dot product of two like vectors is the square of their shared length. That being said, normals will of course be unit vectors, but I mention this for the sake of accuracy.
Amazing video! Do you plan on talking about parallax shaders in the future?
EDIT: Also, how about stencil buffer?
Both of these shaders interest me :)
No plans for parallax mapping in the future, but if you have GDC vault access (or you can get access through a friend/work/whatever), part of the talk "Shaders 102" covered a similar technique in Unreal. The talk as a whole is worth checking out.
And stencil buffer ahhh.... I used it in my See through effects video but it's not the best use case for stencil stuff. Compelling reasons to use stencil buffer are pretty few and far between as most ways to use it can be achieved with the depth buffer already. I'll try and think of a cool way to use it for a future video hopefully!
I appreciate the reply :)
I know these might be really easy but maybe could be a quick video that gets good traffic (because of the game). Maybe you could do a quick case study on the enchanted items icons in and the end portal in Minecraft. And whatever other cool shader stuff is in that game...Can't think of any off the top of my head. Love the depth of knowledge you show on your videos ...i only understand some of it but it reveals there is a lot I didn't know i didn't know. Maybe these (what i presume to be) simple shaders or tricks would be something a broad lower level (relatively speaking to yourself) audience might want to implement and can be knocked out relatively quickly. Also maybe you could consider doing a video showing us the best way to interact with shaders via script such as turn them on/off like in the end of the pokemon dive or modify them with gui controls or through user selection, like you did in your web demo of the Spelunky shader..A unity talk on mobile optimization has me paranoid about mobile optimization..if you watch it you will understand why ahah but he talks specifically about shaders at 34:35. czcams.com/video/j4YAY36xjwE/video.html. Maybe you could talk some about shaders and running them on mobile. Just some thoughts. Probably the most technical in depth unity information with professional content on youtube I've found . Great work hope to see more soon.
"I'll get into art as I recover burnout after studying computer science in university. I'm sure everything will be just very calm and no math involved. :)" (jokes aside really good video but lol)
Can you Please Make tutorials on "Water Effects like in ori and the blind forest , RayMan" it will be very helpful :)
How would one create normal data if they don't have the original objects but they do have a texture? Note: This is a technical question, not a "use application " question.
Awesome video. learned a lot about normal mapping because you just showed it how it is. Best way I learn things :D What is that thing with the top hat? Looks like a state of the US?
I Have an Idea what you could analyse next. Prey's Looking Glass technology... it's basically the same as Portals in Portal but still very interesting and a shader break down would be great
waiting more awesome videos for shaders and effects....It is really cool...thanks a lot (Y)
while i also use unity, it would probably be helpful to also highlight where all these precalculated values like the matrices come from
bitmap texture = matrix
Great stuff. New subscriber!!
Am I just crazy or are you using Donkey Kong Country Soundtrack as background music (5:41 sounds like stickerbrush symphony)?
Edit: Lol just saw the Video notes and I'm right :-)
I would love if you went back to the exercises you had in shaders 101, it helps reinforce what you are saying so I am not just listening along. Keep up the awesome videos though!
Really Great!!
Thank you.
What a well explained video…….What ?
love these
at 4:50 you show the matrix multiplication. shouldn't you do mul((float3x3)unity_ObjectToWorld, v.normal) instead? You use the inverse matrix which also happens to be the transpose on the right side, matrix multiplications are not commutative so while in this particular case it works and it does transforms from object to world space it might not always be the case. If I'm not mistaken unity objects scale does some weird stuff if you get it wrong, besides, its the correct way to read it.
Yeah I probably shouldn't have moved past that so fast. What you've written here will work but only for uniform scaled transforms. Right side by the inverse will work properly with scaling. It's been a while since I've looked at the proof for that, I'll try and find a good source and make an annotation. Thanks!
really? i thought i was the way around, please do share that information when you find it, I might be doing the wrong way all along without realizing
Anyone who understands what your saying doesnt need this tutorial.
you are a GOD!!!!!!!!!!!!!!!!!!!!! THANK YOU!!!!!!!!!!!!!!!!!!
Hey! I know this is a bit off topic to your current video, but do you know of any smart way of masking off the world to only render objects that are inside a sphere? I've tried using the stencil buffer but the problem is that if you angle the camera you will continue to see the object despite of it not being physically inside the sphere... I need to achieve a type of snow globe effect, like a little world inside a sphere☺️ Thanks for the awesome videos! Keep them coming👌🏼
Not entirely sure what you're trying to achieve, but a sphere can be defined by just a point and a radius. So if you're writing/editing the shader of the objects that are within the globe, you could pass through world space positions from vertex shader and use clip/discard with the globe's point and radius.
Question. Question. Question. At 4:50, shouldn't it be the ObjectToWorld matrix??? And also, shouldn't the matrix be the first parameter? Is the opposite order of the parameters what causes the matrix to be the inverse?? So many questions.....
Very good questions! I don't know why I skipped over this. We are transforming a normal, which is slightly different than just transforming a direction (that's what the math you're suggesting performs). You need to multiply a normal by the Inverse-Transpose of your model matrix to properly account for non-uniform scaling. If you look up why you have to do that you'll come across a stack overflow page with several proofs and different explanations, but the goal is to not squash/stretch your normals if you're scaling non-uniformly.
Then a little trick that Unity actually uses in their cg include files, is to perform the multiplication on the right side. This is the equivalent to multiplying by the transpose from the left, except that mathematically you'd be multiplying a 3x1 matrix instead of a 1x3. But shader compilers don't care, they let you use a float3 as a column or a row contextually.
So the inverse of ObjectToWorld is WorldToObject, and multiplying from the other side accounts for the transpose, which all in all gives us the effect of multiplying by the inverse-transpose.
If you can assume uniform scaling, you can just use ObjectToWorld from the left-hand side as you have suggested and your normals will look just fine :)
Thanks for your response. This actually motivated me to spend 3 hours reading up about the whole matrices things. Thanks for being informative as always. Also I just made a video showing off some of the shaders I wrote. You can check it out on my channel if you want. No pressure tho.
Wow there's some really cool stuff in there! The mystique transition is awesome. How does that look on a more complex model?
Makin' Stuff Look Good I'm glad you like it. It looks pretty cool. It supports tessellation so you can effectively swap out the model by baking 2 different models on the same base mesh. I'll probably make a video highlighting each of them. Again thank you for your feedback :)
I love your work dude. I'm putting a bug in .. for fur/hair shader!
that's so much helpful
You should clarify that with the dot product you're only talking about unit vectors, for example (2,2) . (2,2) = 8 not 1
Good thing I'm an engineer. Beat my 60% understood!!
God that into makes me wanna die
Could you make a video on Terrain shaders? One with triplanar for dealing with texture stretching and UV resizing by distance to reduce patterns?( that would be amazing, lol). Anyway, there's hardly any info on terrain shaders for unity in the internet. Most free terrain shaders out there are 4 to 5 years old and don't work anymore. So it doesn't matter if you just want to use them or learn from them, because they just don't work anymore. So I was thinking it would be of great use for many people.
Just my opinion but it would have been better if you used triangle for 3 point plane, as triangle is a figure which can be drawn only two ways given 3 points (the only difference is which way normal is pointing), is why it's used everywhere instead of something like a quad or hexagon, as other figures don't have the same property.
Also it's not hard to calculate normals yourself, even though it's useless if you're using an engine, since it will always have a way because calculation is simple and works for any valid mesh.
There are only three points, but 4 are visualized. The fourth is actually point A reflected across BC, that's why there's no control handle for it and it moves on it's own.
To visualize a plane, what I really wanted was one that stretched out to infinity like a plane actually should. But that visualization didn't read well so I changed it to a quad.
Gamma correction has barely nothing to do with human vision... That's a misconception / myth, and it's making me angry whenever I see it propagated further. Our eyes, just like a camera's sensors, record values in linear space. We see linear values. We need linear values. We never see gamma values.
The only reason for gamma correction is backwards compatibility with old CRT technology. Back then, the monitor tech was so bad that it was displaying darker values than those expected and supplied to the GPU (coincidentally, in a non-linear fashion). Which meant that a value of 120 was displayed as less than 100 (not accurate, but just for example). People figured out that by supplying higher values than originally intended, they could get the monitor to display the required brightness. To do that, they started to gamma encode values to higher values (which again, works in a non-linear fashion). By supplying these gamma encoded values, the monitor was (through the nature of it's cathode ray tube) capable of decoding those values to the required lowered ones (keep in mind that this decoding was an abstract, conceptual decoding. There were no actual mathematics performed by the monitor)
After the dawn of CRT technology and the birth of LCDs, monitors were finally able to correlate the GPU supplied data with on-screen brightness (which meant that 120 was 120 on the screen). The biggest issue, however, was that the ENTIRE INTERNET was filled with gamma encoded images, and backwards compatibility was required (there was no way to purge all the servers and change the standard to no gamma correction without making all images up to that point obsolete). "Why is that?" - you might (hopefully) ask. Well, if you try to display a gamma corrected (encoded) image on an LCD with gamma correction features disabled (gamma = 1.0), you'd get a very bright and washed up image (remember, gamma encoded / corrected values are higher than those intended).
LCD manufacturers figured that the only way to keep going forward was to cancel gamma correction from images at software + hardware level. The LCD has a LUT (LookUp Table) chip which is responsible for doing gamma expansion / decoding, effectively canceling gamma correction and displaying values as originally intended (basically doing, through mathematical means, the same thing that a cathode tube was doing naturally). To enable this decoding process, you need to set your monitor to a gamma of 2.2 (the number was chosen as an average between 2.0 - 2.4, during the old CRT days, because it was giving the best results for a wide range of monitors), because almost every artist is encoding using that value. Of course, panel quality is very important here, so a value of 2.2 might not provide the exact experience intended by the artist (the LCD panel might be bad at reproducing colors / have bad background lighting).
Basically, the entire process is like this: an image is created (values are in linear space) -> then gamma encoded (non-linear space) -> stored somewhere. Somebody comes along and downloads said image -> opens it -> (OPTIONAL STEP) if they want to alter the image, they have to do all their calculations by first decoding the image (getting it in linear space), then calculating, and after that re-encoding (back to non-linear space) -> gets sent to the GPU -> GPU sends it to the monitor -> the monitor uses its LUT chip to decode the image (linear space) -> image is displayed.
Gamma correction is at this point a useless vestigial artifact of a bygone era, on account of bad early tech.
You should do more content, i suspect you are sitting on a goldmine! Love your format btw...
What about a tutorial about making shaders for vfx, like the Diablo Way with multiple multiplied scrolling textures.
Which software are you using for those animations? Your Videos always look so polished
Flash for some 2D, and OBS captures from Unity for 3D.
1:41 i think it should be right hand rule, not left hand rule. or is this different in shaders?
hey nice video but is it possible to explain why more about the blueish purple color being that color? eg. when we divide 255/2 = 127.5 = results in 128 for Red and the Green Channel in the mid value but for the blue when we divide 255/1 = 255 we get a (128,128,255) value. SO my question is is 127.5 = 128 or does the dividing the R and G channels should be 256/2 = 128?
thanks
worldnormal=red*tengent+green*bitengent+blue*normal(blue show how much normal want to use)
So, to be clear... the normal map doesn't contain raw normal vectors. It contains what percentage of the normal, tangent, and bitangent go into the final normal, and the normal and tangent will be provided in the geometry data. Is this correct?
Yes, that's a very good way of explaining tangent space normals! Thanks for the comment ^-^
Then I'm confused as to how moving the light to tangent space works. There, it appears the normal in the texture is the real normal. I understand that the full math behind tangent space is beyond the scope of this video, but it would have been helpful to see some sample values in the normal map, and show what the values are for the normals in world for tangent space.
Hey! You have shared some shaders in your assets - VisualObjectSpaceNormals.shader is the same as VisualWorldSpaceNormals.shader... probably you forgot to delete "UnityObjectToWorldNormal()" from your vertex-shader ;)
Great video, thank you very much! I am waiting for some video about parallax mapping!
Whoops! When I was recording parts for the video I would just quickly change something to get the shot I need instead of changing which materials were applied so I probably goofed. Thanks for the note, I'll fix it up later today!
thx
If using forward rendering, why would you want to transform per fragment normals from tangent space into world space?Isn't it better to transform light direction/position from world space to tangent space instead? It would save operations in the fragment shader.
9:06 - yup I mentioned that. It's generally easier to think about and visualize the normals in world space so I explained it this way. As well, there are effects that require world normals anyway so it's good to understand how you would get them out of a normal map.
Sorry i missed that, yeah world space normals are useful too in some situations.
The normal vectors at 1:50 are pointing in the wrong direction. ab x ac should be pointing downwards.
To be consistent with the rest of the video, that diagram is of a left handed coordinate, with Z point towards us and Y point up. B and C's x values are actually lower than A's.
Check this: imgur.com/J73qvjn
ab x ac is blue, and ac x ab is red.
Sorry for the confusion!
I see, sorry for the mistake.
Hey man, any idea on how Hearthstone make those green energy auras around selected cards? Im not sure if it isnt just animated texture/sprites or if theres an actual fancy effect, or maybe a combination of animation + some gloom.
How do you calculate the normal at a vertex? If a normal is usually found with a cross product, are vertex normals just averages of the normals of the faces around them?
vertex normals come from the input (from the triangle mesh)
Why can’t we just use the normal value instead of having to calculate the tangent value?
@5:09 could you explain this bit more? How did you get that texture? I’d really like to be find a good way to take 3D animations and bake the, in 2d with normal maps for games.
Great video but, wouldn't it be faster and more efficient to use screen space normals instead of world space?
I mention that around 9:00 that you can instead transform the lighting direction into tangent space and then you do a few less ALU's in your fragment code. There's an example of that in the github project.
As for screen space normals, that would be the approach if you we're writing a deferred shader. If you dig into the Unity cg cinclude files they have some good examples of that!
Ah. So your working with a forward render I assume?
Yeah, Unity has some support for Forward and Deferred out-of-box. In this video, we're writing a forward base pass with just one directional light. Things can of course start to get much more complicated as we support more lights/lighting features. But to keep the video from being an hour long, I tried to keep it simple :D
Ah. Are vertex shaders affected by whether or not you use a deferred or forward renderer?
Yes you'd typically write separate vert shaders for forward and deferred because they might need to pass through different data.