NOTE:
Please
be aware that I will not be giving instructions for use with 3DSMax throughout
this document. I expect you to have a basic knowledge on how it works,
how to assign a renderer and set up its parameters, how to create primitives,
how the material editor works, and so on. If you have never used 3DSMax
before, you’re in over your head attempting this tutorial.
What
this is all about
This
all started when I made the following post in the Ideas & Suggestions
section of a forum for GMod, a Half-Life 2 modification:
http://forums.facepunchstudios.com/showthread.php?t=21286
Up to
the time I made this PDF, the thread has received over 600,000 views,
which isn’t what I expected at all. I simply intended to have Garry (the
creator of GMod) consider putting HDRI technology into GMod. Real-time
HDR lighting, however, has already been achieved using this program created
by someone whose name I can’t find:
http://www.daionet.gr.jp/~masa/rthdribl/
Since
I started the thread, I have received hundreds of emails requesting tutorials
and help with setting up HDR lighting in Max. Now, instead of helping
every single soul sitting on the GMod thread, I have decided to make this
PDF explaining how to do everything.
HDR
Lighting explained
Firstly,
a HDRI is a High Dynamic Range Image. It is a single .HDR file composed
of one photograph at different exposures. A few years ago, the special
effects industry discovered that photographing a shot at multiple exposures
is a way of capturing multiple levels of light and that they can be used
to realistically illuminate a synthetic scene. Soon they realized that
by doing this, they could illuminate a synthetic scene using the exact
same lighting conditions as a real-life one, resulting in a very realistic
computer-generated image when rendered. This means that the photographer
not only has to capture all levels of light in the area, but also all
the possible angles which it’s coming from. The way they do this is through
using a 100% reflective sphere. With this sphere, we can photograph absolutely
everything in the scene, and then photograph it at multiple exposures
to capture all levels of light as well. Since we can only ever see half
a sphere, the photographer must photograph the sphere from one angle,
and then from another angle (ideally 90 degrees around from the first
angle.) S/he then uses a software program to merge these exposures together.
The exposures are not lost; they are simply "packaged" into
one .HDR file. That software is then used to "warp" both sides
of the sphere into one, which is then called a Light Probe. This light
probe is warped into a complete panorama, which is then wrapped around
the synthetic scene as a sphere again and used to globally illuminate
the synthetic scene. This is what’s called ‘Image-Based Lighting’, but
not only is the photographer using those images as a light source, but
he’s using all the levels of light which he captured from the multiple
exposures.
In conclusion,
a High Dynamic Range Image is one which contains multiple levels of light
(more than one exposure), while a Low Dynamic Range Image is simply one
image (which is of course at one exposure.) An LDRI can be used to illuminate
a synthetic scene just as much as a HDRI can, but because a HDRI contains
multiple levels of light, the HDRI will result in a more realistic render,
particularly when a synthetic object is reflecting it.
The
orb explained
The reflective
sphere is the number one thing I’m always asked about. What I use is a
Garden Ornament I bought at a Nursery for $15 Australian dollars. It’s
a hollow stainless steel sphere, about 14cm from one side to the other;
a perfect size. You want nothing smaller than a Christmas Decoration and
nothing bigger than a basketball. If it’s too big, it will dominate the
image. If it’s too small, the reflections will not be spherically accurate.
Since I bought my sphere, I’ve done many things to it. I’ve cut a hole
through the bottom and cut a thread through the hole, so now I can screw
the sphere onto a tripod. Because the width of the stainless steel is
fragile and very thin, I’ve screwed a tripod fitting into it instead,
so now I’m screwing the tripod fitting into the tripod instead of wearing
out the stainless steel of the sphere. The next thing I did to this poor
orb was paint half of it a perfect grey. Now I photograph that side too,
and then create a synthetic sphere of the exact same shade and match it
up with the real one by altering the HDR image’s colours, brightness and
contrast. Once the synthetic grey sphere looks exactly the same as the
grey side of the real sphere, you know for 100% sure that synthetic lighting,
colours, brightness and contrast are perfectly matched with the real-life
lighting conditions of your photograph.
Since
I wanted the grey half of the sphere to be an absolute grey (50%
black and 50% white) I went down to Peel Paint Supplies and had them custom-mix
the paint for me. I also got them to put it into a spray can format, because
painting it with a brush would leave streaks along the ball’s surface.
You need to make sure that the paint you ask for is a Matte paint (non-shiny.)
What
you will need:
If you
only have one tripod (for your camera) you’ll have to hold your sphere
up by hand while you photograph it. This of course means that the sphere
won’t be aligned in each of your exposures, so you’ll need an image-editing
program with layer support to line the sphere up in each exposure. It’s
critical that the sphere is in exactly the same position in each
exposure.
First
thing’s first! You must photograph your ball at multiple exposures. If
you’re photographing outdoors, a good range is from 1/4 of a second to
1/1000th of a second with 1.5 F-Stop incriminations. If one
of your exposures is completely black or completely white, don’t bother
photographing it. It won’t contain any information about the lighting
in the area. With this in mind, what you really want to do is take exposures
from almost completely black to almost completely white. The most exposures
you should ever need are about 5 or 6. I’ve managed to get away with 3
before, but with 5 or 6 exposures, you’re sure to get all possible levels
of light (going from 1/4 of a second to 1/1000th of a second.)
When photographing the orb, try not to photograph it from a short distance.
The closer something is to the lens, the bigger it gets, which means the
middle of your ball will be bigger than the rest if you photograph it
from close-up. Get back about half a meter to a meter and zoom in. Like
I’ve said before, it’s critical that the orb is in the exact same position
for each exposure. Nothing should change but the exposure settings
(so you might even want to manually focus on the sphere to prevent the
auto-focus from thinking otherwise.)
Once
you have photographed all the exposures, you should then switch your camera
onto Automatic mode and let it decide what exposure your background (final
render) will be at. Point your camera at the ball and write down/remember
the Shutter Speed and Aperture your camera gives you. Switch back to Manual
mode and apply the exposure settings which your camera gave you. What
I do now is photograph what I call a "Probe Reference Shot."
Using a manual focus on the ball and a manual exposure (which Automatic
mode gave me), I photograph the ball so that it’s only consuming one half
of the photo (left or right.) This way, I can create a synthetic sphere
next to the real sphere, make it reflect the HDR image, and use it to
align the synthetic camera angle until the synthetic sphere’s reflections
are exactly the same as the real one’s. This is not to match colour and
contrast, but simply to get the synthetic camera angle the same as the
real one, and to align the HDR image with the lighting in the real-life
scene.
After
photographing the "Probe Reference Shot", I turn the ball around
so the grey side is facing the camera, and then photograph that (again
with the ball only consuming one half of the frame.) This is what I call
the "Colour Reference Shot". I’ve explained what this is for
in "The Orb Explained", but we’ll come to the reason for this
later in the tutorial.
Now that
you have your exposures, your Reference Shot and Colour Reference Shot,
you may want to photograph a background image for your synthetic objects/characters
to be rendered into. Make sure you photograph your background shot at
the same angle, focus and exposure as your two reference shots. Once you
have all this stuff, the photography side is complete!
Notes
on the photography
If you don’t have a grey side on your ball, try using a ping-pong
ball. Be careful though, white can be easily exposed, especially with
outdoor scenes.
For your reference shots and background shot, make sure the focus,
exposure and camera angle is the same.
Photograph the ball from a distance, about a meter or so. The closer
something is to the lens, the bigger it is in the photo, and you don’t
want a ball with a fat centre.
When you photograph the exposures of your ball, the camera must
be at the same hight (in other words, it should be looking at the
ball straight on and not be looking up or down at it.)
You should photograph your reference shots at the same camera-angle
as your background shot.
If you’re photographing two sides of your sphere, make sure the second
angle is 90 degrees around from the first angle. Try to keep the same
distance from the sphere as the first angle. Also, the exposure values
should be the same as they were for the first angle, and you should
end up with the same amount of exposures, too.
2.
HDRI Assembly: Single sided
Now that
you have all the photographs you need, it’s time to get them onto the
computer and assemble your HDRI. Once you’ve uploaded your pictures, the
first thing you need to do is load up HDRShop. Go to Create > Assemble
HDR Image Sequence and you’re presented with the "Assemble HDR image
from LDR Sequence" window. Click Load Images and select all of your
exposures either from darkest to brightest or vice versa. Once the exposures
are loaded, the common thing to do next is to click Calculate under ‘Calculate
Scale Increments’. What this does is calculate the amount of F-Stop increments
between each exposure. If you already know this value, select or type
it in under the ‘Use Preset Scale Increments’. Once that’s done, click
the ‘Generate Image’ button. You now have your HDR image. Use your numpad’s
plus and minus keys to browse up and down through the exposures. Once
you find the one which is reasonably close to the exposure used in your
background and reference shots, click Image > Pixels > Scale to
current exposure. This sets the currently displayed exposure as the default
one for the HDR image, and prevents you from having to increase the brightness
when you load it in 3DSMax. Without doing this, the darkest exposure of
the HDRI will be the default exposure.
Now we
have to warp our HDR sphere into a HDR panorama. Click Draw > Options
> Circle, and select the sphere in your image so that the circle contains
the sphere perfectly. Now go to Image > Crop, and the image is cropped
to the sphere. Go to Image > Panorama > Panoramic Transformations.
Make sure your Source Image (on the left) is set to Mirrored Ball, and
the destination image is set to Latitude/Longitude. Hit OK and you’re
left with your mirrored sphere warped into a panorama. Notice how the
default exposure you set has been maintained through the warping process.
Lastly, save you HDR panorama in Radiance Format (.HDR).
HDRI
Assembly: Double-sided
Photographing
both sides of the sphere is ultimately a way of removing all distortion
in your final panorama (which appears on the left and right side due to
the very edges of the sphere), and you can also use this method to remove
yourself and your camera from the shot.
Once
you’ve combined your exposures into one HDRI for each angle, you must
now set up points which show HDRShop how the camera has rotated around
the sphere for the second angle. From hereon I’ll be referring to the
angles as Probe 1 and Probe 2. Load up both probes and open the Point
Editor (Window > Point Editor). Go to Points > New Point, and select
a feature in Probe 1 which you can also see in Probe 2. Once you’ve selected
it in Probe 1, you must also select it in Probe 2. Create a second point
in the Point Editor and do the same only with a different feature visible
in both probes. Once you have both points assigned in both probes, open
up Notepad and type the X and Y co-ordinates for the points for each probe.
In Probe 1, go to Image > Panorama > Panoramic Transformations,
and set it up so that your destination image will be a Light Probe.
Click Match Points and then Settings, and type in the co-ordinates you
stored in Notepad into the appropriate fields. Once you have the co-ordinates
typed in, make sure the first image is set to Probe 1 and the second image
(on the right) is set to Probe 2. Hit OK and generate the Light Probe
of Probe 1. You now have an actual HDR Light Probe composed of both HDR
angles. Save it as Probe1_warped.hdr.
The next
part is optional; getting rid of the distortion and the photographer in
the shot. Go to Probe 2 and transform it from a Mirrored Ball into a Light
Probe. Save that as Probe2_warped.hdr. Re-open both Probe1_warped and
Probe2_warped so that the file names get registered into HDRShop. What
we have to do now is open up an image-editing program with layer support.
For this tutorial I’ll use Photoshop. In Photoshop, open up Probe2_warped
and put Probe1_warped as a layer on top. What you have to do now is create
a mask. Right click on the top layer (Probe1_warped) and select Create
Layer Mask. Using the Paintbrush, paint with black to ‘erase’ the parts
of Probe1_warped (of which Probe2_warped will show through) and use white
to restore any parts you may have accidentally painted. Once you’re happy
with the way everything looks, hide all the colour channels of the document
apart from the Mask layer, so that only the mask is showing. You should
have an image that looks something like this:
This
is your mask. Invert the image (Image > Adjustments > Invert) and
save it as a BMP file called Mask.
In HDRShop,
make sure you have Probe1_warped and Probe2_warped loaded, and then load
the third element: the mask. Go to Image > Calculate, and select A
* C + B * (1-C). Set Image A to Probe2_warped, B to Probe1_Warped, C to
the mask, and D to Probe2_warped too. Hit OK. HDRShop merges Probe1_warped
and Probe2_warped into one using your Mask.
This
concludes the assembly of your double-sided light probe, and what you’re
left with now is the one and only HDRI which you will use to light your
scene.
Notes on HDRI assembly
Image quality is not an issue with HDR images. Poor quality is something
you’ll get naturally when warping your sphere into a panorama, and is
not avoidable. If you could see the light coming at the sphere as an
image, you can imagine that it wouldn’t be a very sharp image at all.
In fact, the whole image would be one big blur. You’ll be blurring your
HDR image in 3DSMax anyway, because this speeds up the rendering process
while also improving the image quality of your render.
I usually increase the exposure by 4 increments each time and select
‘1F/Stop’ in HDRShop as my exposure increment value. This always gives
me perfect results.
3. Setting up in 3D Studio MAX
Now that
you have your HDRI all ready to go, it’s time to set things up in 3DS.
Once you have 3DS open, the first thing to do is set up the environment
and Viewport background.
- Go to Render
> Environment, and load a Bitmap into the environment channel.
The bitmap you want here is your Probe Reference Shot (the reflective
side of your sphere), because you’re first going to line up your HDRI
and adjust its brightness and contrast to match the sphere in your
Probe Reference Shot.
- Go to Viewport
> Background and tick ‘Use Environment’. Close that, right click
on the Perspective viewport’s title and tick ‘Show Background’.
- Create a sphere
from the middle of the grid, about the same size as the sphere in
the Probe Reference. Rotate the view so that you’re looking at the
sphere from front on (like your camera was looking at the sphere.)
The grid should now be appearing as a horizontal line. Adjust the
view so that your CG sphere is next to the real sphere in the viewport’s
background.
- Once your renderer
is set up, drag the HDRI to a free material slot in the material editor.
Make sure it’s set up as a Spherical Environment. What you have to
do now is create a new material for your CG sphere. Since you’re matching
up the reflective side of your real sphere, create a new material
which is completely black and reflects your HDRI 100%. Simply drag
and drop the HDRI as an instance to the Reflection channel
of the material. It is very important you drag it as an instance.
This ensures that whatever changes you make to it in the material
editor, they’ll be linked and updated in real-time to where you dragged
the HDRI from. Your material should now be reflecting your HDRI and
should be looking something like your real sphere.
- Drag and drop
the material to your sphere in the viewport. You’re now ready to give
your scene a test render and see how close your CG sphere comes to
the real thing. If it looks like the reflections are the wrong way
around, rotate the view in the viewport to correct it. It’s better
to rotate to correct the angle than shift the HDRI in the material
editor.
What
you want to do is adjust the RGB Level and RGB Offset of your HDRI (under
the Output rollout) to get your CG sphere’s reflections as close as
possible to the reflections of the real sphere. The RGB level is the
intensity level; think of it as the Brightness & Contrast levels
in Photoshop rolled into one. The RGB Offset is your brightness level.
To increase the contrast of your HDRI, for example, you would raise
the RGB Level and lower the RGB Offset. This means having to give your
RGB offset a negative number, because it starts off at 0. If you increase
the RGB Offset and leave the RGB level as it is, the HDRI’s brightness
will increase but the contrast will stay the same, resulting in a "hazier"
image.
You
may also want to use the individual colour options for your HDRI. To
do this, tick ‘Enable RGB Controls’, click R, G or B, and adjust the
corresponding the line. For best control, use the highest point on the
right. The shadows are obviously on the left of the graph, while the
highlights are on the right, but by having just one point on the left
and one on the right, you can adjust one while adequately adjusting
the rest of the line at the same time.
-
Once
the colours and contrast of your sphere’s reflections are looking
the same as the real ones in the Probe Reference, you now want to
further match the colours and contrast using the grey side of your
sphere. In the environment window, change the Environment to the Shade
Reference Shot (the one of the grey side of your sphere). Assuming
it is of 50% black and 50% white, create a new material in the editor
which is of the same colour (adjust the grey slider to the middle,
around 130 or so.) Apply this new grey material to your CG sphere
and render it out. Now, if your CG sphere is darker/brighter than
the real one, try not to change the RGB Level and RGB Offset of
your HDRI. Instead, change the intensity value of the GI/Skylight/Direct
Light which is using your HDRI. This is because you’ve just matched
up your reflections perfectly, and what you’re doing now is adjusting
the lighting, not the reflections.
-
Finally,
once your CG reflective sphere and CG grey sphere are close to the
real ones, you should now have a good lighting set up and should end
up with a realistic composite when you render something into the background
shot. Change the Environment to your background shot, and rotate/zoom
the camera to match the angle of the real one. You now have a HDRI-lit
scene ready to go. What you do hereon is up to you!
Notes
on Setting up in 3DSMax
- It is often a
good idea to have two copies of your HDRI sitting in your material
editor; one which illuminates your scene and the other which objects
in your scene reflect. This is because a sharp un-blurred HDRI used
for lighting can often cause your render to look dirty and poorly
sampled, while a blurry one can improve this greatly without changing
the lighting of your scene at all. Not only that, but a blurry HDRI
can actually speed up the rendering process, because there are less
fine-details in the HDRI to be processed which make no difference
to the lighting of your scene anyway. Only create a duplicate of your
HDRI once you have your lighting, colours and contrast set up and
matched to your liking. Otherwise you’ll be changing settings for
two HDRIs instead of one should you decide to change them in the future.
To blur your HDRI, increase the value of the Blur Offset. 0.02 is
often a good value; it blurs things nicely while retaining details
in the HDRI. Rename that material as ‘Lighting’ or something similar.
The duplicate HDRI (which you should rename ‘Reflections’) should
remain sharp to have nice sharp reflections in your scene. The Reflections
HDRI is of course the one to use for your object’s Reflections channel.
4.
Creating Matte Objects
Whilst
compositing CG elements into a real-life photograph, you will most likely
come across the problem of casting CG shadows onto real-life objects.
The way to do this is to create Matte objects. A Matte material (called
‘Matte/Shadow’ in 3DSMax) is one which is invisible to the camera but
shows only shadows cast upon its surface. In other words, we can’t see
the object with that material applied, but we can see shadows/GI
cast upon it. Here’s how to set up a Matte surface for Mental Ray, Brazil
and VRay.
For VRay:
VRay comes equipped with its own Matte feature, so you don’t have to use
MAX’s at all. Right-click on the object you want to be a Matte, and select
VRay Properties from the menu. Tick ‘Matte Object’ and ‘Shadows’, and
you’re all set!
For Brazil: Brazil’s method is a little more complicated.
For Brazil you have to use MAX’s Matte/Shadow material, but this doesn’t
cut the butter just yet. If you left it like this, your background image
would be the material which the Matte object would take on, so when it
bounces/generates light onto other objects in the scene, those objects
will appear translucent. To resolve this matter, we’re going to set up
a Brazil Utility material. Assign a Matte/Shadow material as the Utility’s
primary material, and assign a material (with the same colour as the real-surface
in your background photo) to the Utility’s ‘Generate UI’ material. This
will ensure that the Brazil Utility will generate/bounce light of that
colour and not of the background photo. Now just drag and drop the Brazil
Utility to the object you want to be a Matte.
For Mental Ray: With MR things are pretty straight forward.
Simply drag and drop a Matte/Shadow material to an object. Mental Ray
isn’t capable of rendering GI shadows onto a Matte (areas which light
can’t get to,) it can only render actual shadows cast by a light
source (such as an Omni or Spotlight.) This is one limitation of MR which
tends to steer people away from it.
Conclusion
So where
to from here? It’s up to you. People often want to get everything exactly
the same as I do, but this is a form of art, so you must experiment and
develop your own ways of doing things. There is no right or wrong way
of doing what you want to achieve. Search the net for more tutorials and
try things out. That’s how I started and it didn’t hurt, look where it’s
got me today.
|
