Transcripts
1. Welcome to the 3D and AI Workflow: Combine through D and AI. This is the point
of this tutorial. I will explain how install stable diffusion at
the beginning level, how to plug in different
modules and how to set up a simple sN in blender that will work in combination
with stable diffusion. Every software we are
using here is free. You just need a recent
gaming laptop, ideally, three or 4-years-old
because we do need an video card to run it on. In this particular
case, I want to create a showroom for a prop, and you can apply this
to any idea of your own. It doesn't have to be
like a medieval room. Like in my case,
it can be Sci fi. So let's start and
enjoy the course.
2. Installing Stable Diffusion Web UI: Go and install step
Body fusion web UI. So for that we need
to go to github.com, automatic 11 lemon Steel
Body fusion web UI. Then we can scroll down to see the installation
instructions. For Linux Windows and
Apple, I'm using Windows. So we need to install Python
and we need to install Git. If I click on Python, we just need to scroll
down to the bottom, find the Windows installer, standard installation,
nothing custom. Same thing for the Git. I click on Link here, and we go just go for
Window 64 beat setup. Download the executable and follow the standard
instructions, click Okay all the way. The next thing I
need to do I will go to my C drive and I'll right click creates a new folder for our stable diffusion. Then I need to click on the URL section and type
in CMD, the command line. Then I need to go and copy
this line right here. Get Clone down to the end, Control V, paste it right
here and press Enter. That now is going to run and download all the files that we need. Now we have the folder. We need to go and
run webpiuserdt pad. Sub click on that. And it's also going to run
through commons here. I download some packages. Don't worry about it,
give it some time. After waiting for a few minutes, and it can take you ten, 15, 20 minutes because
downloading a lot of files. We got the webbi opened. It is, however, looking white. So if we want to make
it a cool dark mode, we can do it pretty easily.
3. Switching to UI Dark Mode: To go back to our main folder
with stable diffusion. Click, right click on webby and user dot
BD and click Edit. Now we have this note pad where I can just
type in what I need. That is dash, dim space
controls to save it, and let's run it again. Just double click on
Webi user Dot BD. And there go, we
have the Dg Zim. I can Control tab, even switch between them. I have two of them
open right now, but I will carry on with
D deemed automatic 11 11
4. Stable Diffusion Standard Models: Unstable diffusion, you
may or may not have this model already
preloaded version 1.5. However, at this point of time, while recording is tutorial, there is already a new
model which is 3.5. It is hanging phase, and you can see this link. You do need to
create an account to download this model,
and I've logged in. Then I need to switch
to files and versions, scroll down to the
list of files. And here I can see this file dot save TensS which we
need to download. There are multiple models currently available
for download. And see 3.5 L Large
Turbo and medium. Medium is the best for
consumer hardware, so you'll still create your good results and will
not kill your video card. Also, medium model is smaller. It's at five gigs, while large Turba is 16 gigabytes and regular
large is also 16 gigabytes.
5. Adding Custom Models from CivitAi: Model, I go from website
called Kviti or cbi.com. Here, they post a whole bunch of models created by the users. They're all for free. You can
pick any and download it. I actually went and picked Epic Photogasm and then
I just press Download. We need to go to stable
diffusion folder to models, then pick Stable
diffusion and paste the downloaded model
right into this folder. You can see the extension
called safe tensors. That extension is
identified that it was virus checked and there is no dodgy stuff in the file. I'm back at table decision, you do want to click on refresh, so that it will go and
scan the models folder. I already have NASA model
I don't load before, and then we have
our epic photogasm. Every time I press
on a new model, it takes some time to download it and update it in the WI, usually a few seconds,
but if the file is big, it can take 40 seconds.
6. Creating Your First AI Prompt: Finally, we can do
our first prompt. C type in like a medieval room, click Generate a super simple prompt without
much detailing. We will be introducing
a lot more. And it created our image.
7. Customizing Image Save Paths: May generate an
image, it gets saved. And if you want to change this path of saving the particular folder
you want to use, you just copy paste it here, and then click Road UI and
you will get it updated.
8. Installing ControlNet for Precision Renders : Image. We can't
really control it. We can change the prom,
but we'll morpho gain something else to
control the output. We need to install Control. Let's go extensions,
click On available. Then I click to Unload From and I got a list of extensions that are available to be
downloaded and installed. Press Control F, G search
for ad web Controlt. And I just push it down, find the second search result of SD WebUI Controlt
and here we go. We have it and I
just need to click I
9. Building a Room Blockout in Blender: Showroom. So inside the blender, 4.3, slightly different from previous versions of blender. I go Shift A, and I will create a cube. So I want to get a cube. I want to make it a
real proportions. So two by 2 meters
is about right. I just want to scale it out, say two by 4 meters, type it in. And then press step to
go inside, Edit mode. I'll press three, which is my
hot key to pick the faces. We can also pick the faces
or variouss at the top. Then I press delete to delete and pick the
faces to delete. So now we have our shell, which if I pick it and move it away can
create a bit more depth. And then I will go Shift A, create a new cube, and I
want to decorate our room. So I'm pressing G to move this, pressing S to scale
it on different axes, Shift D to move it across. Shift D, shift D. And
I want these guys to to fill up the room so
they go a bit more infinite. I might consider to add
a bit of ceiling here, so I will go picking the
edges click that one, Ship, click this one and Control
R to create a new age loop, and I use my middle
mouse button, scroll and add a
bit more divisions. And then I pick this guy, middle guy and go with
the proportional editing, I can now push
this all together. If I press G and use middle W, I can scroll up and down and press Z to go up in
the axis that I need. So I'm creating this
little arched ceiling then I press one
for tic selection. I actually want to go
into transparent mode, so then I can pick all the visible and
invisible vertices, press G, move it down, and I have my cage by accident, I hit on the floor. So again, I here press to enter the wif re mode and pick all
the vertices at the bottom, tone up portional editing and G and Z plus zero to
scale it down to zero. Now I have a flattened floor. Let's go back, turn of
the transparency mode, which helps to select the
back faces back vertices. And I will just shift
D plus R, rotate, shift rotate to
snap the rotation, press S to make it bigger
again, to scale it up. And I want to create
another pot there, press S and Z to
scale it upwards. We create a little very
minor decorations. Something there because
we want to give AI ability to think a
little bit for itself. One thing I want to
add, I want to add a bit of bubble to
all the shapes, press Shift click on
each one of them, press stab and everything
is already selected. So go Control B to add a bubble. And I just add a little
bit of smooth beveling. We can see that beveling here
wasn't happening accurately because these object has
been scaled on one axis, so it's not like a
uniform scaled object. And to fix that, I
have to basically, let's go do this press step and call Control A, apply scale. I will apply the scale transform
and will fix the scale. So now it's a uniform 111 scale press step
called Control B. And now I can get a bit of these softer shapes because cell body fusion does prefer
to deal with softer shapes, and I feel like projection might look a bit
better if there is some wiggle room for when we start doing the texture
projection with cameras, it will be helpful to have
slight softness here. I think these guys, well, it looks a little bit
too jacket that line, so I can just pick all
these guys together, and I will bevel and I'm using my middle wheel to get
around amount of divisions. So I've got a few
divisions running might have nice curve. And maybe I'll just add this columns at the end over there and move
the skies across. Instead of recreating
that on the right, I will just move it and
duplicate this model there. I think maybe it doesn't
have enough depth, I would just go and scale it in one
direction and Control A, apply the scale transform.
10. Setting Up the Camera in Blender: Want to position my camera now. So was on my camera it's
somewhere there's a default one. I will press G. Is by pressing, I'll shift every time
I'm rotating and I'll shift snap snaps to
the viewpod left, right, top, and front. So let me rotate
this guy around. Actually, let me go into
these settings here, just click and drag
and type in a zero. So now I know that my camera isn't under some
kind of funny angle, press, press control,
snap rotate, and we have a to make sure we
are in the right position, I have to go into
view and camera, set active object as camera, and then go into the view
and look camera to view. You can see the part, you know, where we
see the room is pretty tiny, that camera gate. So if I press home, it will bring this camera
gate to the full screen. So then I press, you know, rotate my middle
mouse wheel shift to, you know, just position this camera in the
best possible way. I think I want to change
my camera settings. So click here on
the little camera, and I want to change
the poco lens to something a bit more
dramatic like 24. So that gives us a bit
more Breathing room. We can even go 18. This is basically
how they photoshoot all those properties or all
the real estate properties. They like to go with super
wide angle lens and make everything look much bigger
than it actually is. So you can play
around with the bells and find one you like.
11. Configuring Depth Pass Renders in Blender: A little preparation for
our render settings. I'll go into output
and put 50% here. It will double down
the resolution. And the lower solution, the faster it renders in
stable diffusion. So it's quite important. Then we want to render
out a depth pass. The depth pass will show the nearest objects towards
us in white in wilt pixels, and the furthest objects
away in black pixels. It's also can be
called a Z depth. So here in data, I just click here on Z
depth to get rendered, and then we can switch
to compositing. I go to Compositing tab. You probably will
not see these notes. You have to click on use
nodes for them to show up. Let's go click Render Image. And right now, we can't
really see anything. We don't actually have
the lights on there, and it really doesn't matter. So, press Space Bar,
look for viewer. We know that will show
our render from before. And we have our depth path. If I go back to our view layer and I click on Z and click on Z, you can see the depth path
appears and disappears. So I can just go drag it down, and it will give us
a little preview. Right now, it's
all totally white. To fix the image, spacebar, look for the map range. Drag it over our line, insert it, and then we have
to play with the values. Basically, I go and direct
this slide and try to find some value that will give enough information for stable
diffusion to work with. Not a bad idea to go
into the render settings all way down to color
management and transform. I want to change it
from filmic to raw, that gives us more
bit of gradients and more gradient variation can be quite useful for
stable diffusion. Sometimes stable diffusion can interpret too much darkness and play has been
really far away. And in that case, we can play
with the here and we can change the contrast settings,
low contrast, medium, bear in mind that even though it was a human eye can't really
see the detail too much, stable diffusion will
be able to pick it up. But I will go with
just none for now. Let's connect our map range to the image and then
go render the image. So there we go, we have our final render and
we want to save it. So image, saves and I create
a new folder called PNG. I save it at 16 bits, and I want a zero compression. We don't need any
further compression. And let's call this like a room, depth path, zero, one,
and save this image.
12. PhotoPea (Free Photoshop) for Gaussian Blur of the Room: It's a little P tip. I'm
actually using Photoshop, but if you want to
use a free software, there is a Photo P. It is an
online copy of Photoshop, and you don't even
need to login. Click here, like
open from computer, look for my render,
double click on that. And what I want to do is blur it using a
negotianbl that will add a bit of suggestivity
for the render for stable diffusion. It
really, really helps. So I'm just doing
like two pixels, 2.5 pixels of smoothing. And we can do different
amount of blurriness and try that all later
in stable diffusion. And then we can just go experts. Experts PNG gives you the window how to save
it and just press Save
13. Fine Tuning ControlNet Depth Settings: Time to have fun. The
tedious installations and blockouts are over. First of all, we want to drop our image that we've
created in Blender. Just click Dragon drop here. Then I want to change
the resolution. Otherwise, it will try to
render it as a square. So it's 960 by 540. And by the way, the
image I've dropped, it's not a smooth image, just a straight
expot from blender. Then we can type in a prompt. I will say something very simple medieval room
with window in the back. And before we click Generate, we have to go and click Enable. Enable of the control net, otherwise it will
not do anything. And then we have to change the preprocessor to depths Zoi. Depoi in my experience
showed the best results, so I will stick to it. Other depth processes can
give some bizarre results. Then important
options to consider, and we have quite a few options, but we only need to
play with a few. I want the control weight
to go down to 0.7. That gives a bit more creativity to the prompt to the generation. If it's one, it will try
to adhere to the stuff it sees and can sometimes
give very ugly results. Then I want to
change seed to ten. That will mean that it's
not going to generate a new image every time
I have the same prompt. And let's just click
here Control type to
14. Generating Your First AI Image: And let's just click Generate. Sometimes it will take
you only a few seconds and sometimes it
will take longer. You can pull up the pump window from the bottom bar and see
the progress for yourself. It's advisable to do a
computer reset to run it. It will run on a fresh
restarted computer, it will run faster. But it all comes down to your video card and how much
VRM you have. And here,
15. Control Depth Glitch Depth Preprocessor Switch: I got. I really don't like it. A final thing happened. One point, for some reason, de processor changed to metas. And Midas gives weird results, so I always switch to Zoi. It's a good thing to point
out if you start to get something weird and
not controllable, maybe the depth
profile is wrong, deps processor So here
I've put another prompt. I'm getting somewhere.
I'm getting somewhere.
16. Crafting Effective AI Prompts: Result I found satisfactory. You can see my prompt festival says medieval dungeon room, black, simple gothic window. It wasn't simple, it would create like four
or five of them. Then broken walls and floor, wooden ceiling
with wooden beams. If I didn't put wooden beams, it sometimes wouldn't
generate the wooden ceiling. And for some reason, 80
millimeter lens did help with the look and made it look more realistic versus non
realistic before. I don't know why it helps,
but that's the way it works. Then I used an image that
was even more blurred, so I blued it to like six, seven pixels on
Goshen blue scale. See here, I have the
control weight set to 07, gave me a good amount of
detail and originality.
17. Embedding Prompts in AI Generated Images: Talk about recreating the
previous results, right? So every time you
create something, you want to save
that information. If you have to go to setNs and go to saving
images and grids, and right now we
already have it. Add extended Info, see
the prompts to file name to exf information
of the file. Then click Apply setNs and now every time you
generate an image, it will be saved with
information in the file. And to retrieve
that information, we need to go to PNG Info. So drag and drop one of
the images I've created, and you can see immediately
we have our prompt. Then they say send
to text the image. It recreated the
resolution at the prompt, but not the control net. So the control we still
have to redo by hand, which is a little bit annoying, but that's the way it works. I have to drag and drop
depth image again. Make sure you don't
make a mistake here. And then if I run generate
and you know what? I actually did a mistake, the depth was set to met us, so it creates us quite
a different image, kind of cool but weird. Let me go change depth
to ZoiGenerate again. And with that, I got
the image I want
18. Upscaling Images with Stable Diffusion Latent method vs Swing: Of the resolution, we
can run an upscale. So here, if I click on Enable
has fix and nscroll this, it will double our resolution and write different upscaler. So each upscale gives
us different results. I'll be using just latent. There is also swing for
IC which I try to use. You basically just
go click Generate and run a double resolution. Require a lot more memory
on your video card. I do have to do a
clean computer restart and then run the generation. Advice it will say,
you'll tell me that I'm out of memory
on my video card. Here, I have my
latent upscaling, and I have swing upscaling. So the detail is
a bit different. You can see it
actually recreated the image and changed
a lot of details. If you want to keep
more of the details, you need to change the
denoising strengths. So I'll go to five. Again, if I drop it too much, it will give me somewhat
of a weird results. And also, it's much
longer, right? It can take five, 10 minutes depending on
your system to generate an upscale image versus just
seconds on a low quality. So here I changed the denoising
settings from 0.7 to 0.5, and it gave me a
very different look compared to this one, 0.0.
19. Quick Online Upscaling with Krea AI: Like to upscale
using creat AI tool. This is a free tool with
a limited use, right? So we can upload, say, five, ten images and run upscale on already upscale
stable diffusion. So we do need to sign up
and create an account, then go to enhance and let me just drag
and drop our image. Now on my left hand side, I have the image I put it. I cannot do two X generation. However, I can do just one X, and this already will
improve my scene. Click on the button and
I just have to wait. And now I have this
before and after son, I'm not too sure that this
new creation is better. I think actually works, but we can run the
different upscale and I can go upscale methods
and say clarity, I can drop clarity resemblance and run enhance at this setting. I think new variation is better. Still not too
convinced if I want to trade the old version
for the new version. Some things I like,
like the ceiling, maybe the windows got
a bit more definition, the floor is interesting, but the side walls
are pretty bad, and we're losing that warm that I'm not too happy about it. I think it's
important to mention this tool if you want to upscale your stuff and to
add some details because it can work pretty
well in certain cases. AI especially useful if your computer cannot do
upscaling on your own. So, for example, here, I'm
dropping my 1,000 pixel image into the upscale and
getting my two k render out. And I think it's actually
doing a pretty nice job over
20. Showcasing AI Image Experiments: I want to go through gallery of failed tests just to show
that it doesn't involve a lot of trickery and
prompt engineering and settings engineering
and changing the settings to
get a nice result. Sometimes I forget to change
the depth from Midas to Zoe. If I don't put a window
inside the room, the lighting goes to
look quite weird. So Immediately see here, put in the prompt to
put a window behind. And the lighting
got from kind of fake d one to a
lot more natural, very natural realistic lighting. Then I was trying to achieve a wooden ceiling and also plane with
different settings of a control net
from 0.7 to 08 and constantly change and change and change you know,
control net goes up. It's less variation.
Control net goes down. It's a bit more freedom. And I was just trying to
get that realistic look. Some of these are not too bad, and especially
considering the ability to do the upscaling
in other software. I can drop it and create AI and upscale this pretty
nicely and then reuse it. Sometimes you'll get
some weird holes, like you can see on the
right, is a black hole. It's bizarre. Oh I have
this weird gis gish window. One thing you can see
though, the moss, I actually did prompted a
mossy flow in the prompts. Then I did mention a castle in a different prompt and I got
a castle in the background. Just, you know, the
left and right walls, they look very like
a cheap game type. So other prompt tests here. Sometimes playing around
with the number of windows. And here, I jumped back to stable diffusion model
from epic photogasm. It gave me this result. However, I mean, it's
kind of looking cool. But at the same
time, it was less controllable than
the epic photogasm. So I switched back
to epic photogasm. And here, I think I
was trying next day. This is my depth
switch to Midas, and I forgot about
it, so I had to switch it to Zoi again. I was also telling the
windows like simple window, medieval window, gothic window, a simple gothic window. It all came from
prompts and variations. You can see right
now, I have a frame. I don't know why this
frame started to appear. To remove the frame, I
started to put a frame in the negative prompts, but
eventually it went away. Like putting 18 millimeter lens helped a lot with the
realism of the scene. Again, I don't know why we black box there when
we play with that. So failed prompts
here again, right? And then I said to improve. I try try try. And it's suddenly look
worse, looks better. Very often, it has to do
with the control net steps, 07, 08, 06, 05. I try all I also tried
blurring the image. I did blurring the depth image that came from Blender a lot. B five pixels by ten, by 15, I kind of
blurring it a lot. I was even adding some noise, like monocrobatic noise to
see how that would break it. It can help. It does indeed help
stable diffusion if the edges of the image
are not too consistent. And of course, it will
be a problem if you're trying to do rendering animation because then it will start to go and be wage from
frame to frame. But in our case,
we're just trying to get one nice render out and our animation will come for free in blender at
real time speed. Here, I don't know why I got this just black line through it. I just had to change
control net and regenerate. I did like two, 300
prompts generations. One of the reasons I was
doing them like tutorial. I could also, photoshop
some stuff, right? I can merge few variants
together to get a better result. However, I wanted to show
in this tutorial how to get just a actual export
without retouching anything. But retouching is always an option, so don't
forget about it. This was a rather nice dungeon. Wooden ceiling was
giving me trouble. It often didn't want to
create a wooden ceiling. I had to say, specific
wooden ceiling with beams, and then it would introduce
a full on wooden ceiling. There's some nice examples here. And then I said, like
landscape mountains and started to get these
mountains in the background, or I said the white sky and painted all the
walls to be white.
21. Projecting AI Images in Blender: Back in Wanda, I pick an object and I want
to create a material. In your case should
be material 01, but in my case,
it's already done a few materials, so it's 03. And then I pick all the
objects, except for this guy. So this guy control dyslecti
has to be the last one. And then when I click Control L, I say link materials, and each object now
has the same material. And if I apply
different texture, it will be applied to
all of these objects. Want to change the surface
material to background. So that background material
is very, very simple. It doesn't have
any shading to it, no specularity, no shadows. And that's what we need
because we already have all our lighting happening on the image on the photo that we
generated in stab diffusion. Go into shading. I already
shows our background material, a space baaimageimage texture, and then go look for
our final render. So I found it, then
I'll just drag the color to color output
in the background, and now we have our projection. Our image looks very dark, so we need to change
the slur space. I will go into the render, scroll all the way down to color management and
change the robe which we had used for the depth to
AGX to get nice render. Can see certain issues
with the projection. It's all skewed and
such. We can fix that. For that, I need to pick
all my meshes again, and we need to add
more geometry to it. Right now, the geometry is
too low for good projection, pick that stuff up
and rise A and then subdivide and do it again and do it again
and do it again. Now it's getting a bit heavy, that should be good enough and project the UVs from you again. I have to go first set view to camera and then
pick everything else, press step, press A
to pick everything that's already picked and
go UV project from you. That will restore the issues. If I press home, you can
see now and overall, I do see the back window, right? The back window is
wobbly because there's not enough geometry on
the back to fix that. I'm just picking this room
going into the invisible mode, pass through mode.
Lead all the faces. I will just use one
of these boxes. We'll stretch it out. I do need to project
from my camera again, press Step, press A, and project
22. Using Sketchfab to Find 3D Models for AI Background: Room to create a showroom
and to find an object, you can find them on many
different platforms. I like to use Sketchfab
because here we have a very, very nice three D preview. They have a very
large library of everything I mentioned about
there. There's also a shop. You can just look for a
medieval statue and there are a few museums here on Sketchfab that upload
the stuff to use. A lot of these models are actually downloadable.
Some of them are not. You have to go and look at
the bottom information. You can see here,
download Tree model, pretty nice model to
download and use. I looking for a while,
I found the statue of St. Andrew or some
kind of scholar. And you can see it mentions the license it in public domain. Let's just go download
it. You do need to create an account if you want to
download these models. I'll go pick my file
format and we obj. I will also include
the textures. I got the archive, need to right click and extract it here. I can even delete it now. Then I go inside the source,
and there's another zip. I unzip that. And also,
check the textures. The textures haven't
been zipped.
23. Importing Model to Blender and Material Set Up: Side Banda file input OBJ and we pick our OBJ.
Move this around. Looks like it's not exactly in the right position
or right scale, so let me rotate and put
it in place as to scale, move it around there. And we do need
connect the textures. I'll go into the shading,
go look for image texture, Control C control V, Control C control BF three image textures
that we need to find, go open a color texture and I can hook it
up to base color. And immediately we can see we have our texture on
our roughnes texture, which will be responsible
for all the reflections and reflectivity of the surface, which I want to connect to roughness and then a normal map, which will create more surface detailing on this
object, this one. Do you need to hook up
through a normal map node. This one going to color. And then this one will be
connected to normal Map. And don't ask me why
we have to do that, but unfortunately we do. I want to do some color
matching because right now it looks like the statue
doesn't belong to this room. It has to kind of receive the colors that are
bound in from the walls, from the window,
all that kind of pinkish, warm yellow light. And I can just introduce
a hue and saturation note that and allow us to
shift the hue into, say, slightly more pinkish and I can drop the situation and I can
drop the value to make it Daka it can take some time to get combination of
all these values to get the object
look like it belongs. We don't have the
shadows right now, so that makes it
look a bit fakish, but we will deal with
that in a second. This is good for now. I feel like it might need
to be scaled a bit more.
24. Matching Lighting to AI Backgrounds: Create a light. Say,
I'll go for point light, place it over there, and excavate the IV IV studio
starts to add some light, and I can't imagine it to
be light from the window. I'm just playing
with the settings, the intensity of the light. And I'm just pressing
one digit right one, one, one, one, two, two. I don't like to press
one and 00 because then my hand has to travel through the whole
keyboard to do that. I can switch point light
to aerial light to give a broader
highlight on the back, create a rim light
coming from the back. And then to fake our global
illumination from the room, I need to create another light. So this one will be coming from the front and it
will be much duller. And then I do want extra
stuff on the face. So I can just bring
that point light that was created by
default with the scene. I'll just shift right click on the body and then shift
a selection to Coso. It will jump over to the three Coso which I
placed. Put it on the face. Obviously, right
now, way too strong. 11. 11 actually is pretty good. The lights are all very white. We do want to make
them a little bit, like yellowish, pink to mimic
that we have in the room. Same thing goes to the front
light, front, big light. We go and do this pinkish light. And also change the
back light as well. Can also use the color
picker and pick from behind. Pick this kind of be coloring. Looks like exactly
what we need. Now
25. Creating Realistic Shadow Catchers in Blender 4-3: A shadow catches, create a plane first, make
it rather big. It's about right
positioning and going into shading and
sign a new material. I don't really need this
principal BSDF here. T delete and I will look for diffuse BSDF without
any extra information. That doesn't mean don't
need and transparent SDF, we want to mix
these two together. I will mix shader, plug it into the surface, and transparent
goes to the first one and diffuse goes
to the second one. We have two of them
mixing together. If I go and put the factor at zero, it
will be fully transparent. If I go put the factor at one, and it will be a
fully diffuse color. Now I want to create a
mask for our shadow. I'll create another diffuse
BSDF then shade it to RGB. Connect these two together and plug it into
our factor node. And then if I just change
little settings here, I'm in the material
settings on the right, and I need to change random material random
method from D to blended. That gets us our shadow, but it looks like we probably need to invert this. All
right. You know what? We have to switch the
inputs, and there we go. I can go up and down, and I have my shadow. Want to undress the
blackness of the plane. So my mask here not
actually white, so I need to make it really
white will lighten this up. We do, however, have this
particular edge right there. So that edge it's tricky to remove because
the geometry there, if I start to push
this plane down, it starts to hit the
geometry that's down there. What I want to do,
I would just go R and press X for the Xs. I'm just rotate it a little bit. I press shift to slightly move
this shift to slight move. And then we get a bit
of intersection there. Okay, R X shift
move a little bit. And I guess getting less of it. You can also type in the bells, but I'll try to.
I want to do it. I want to eyeball.
Other thing I can do, I can try to play with the
positioning of this light. And this is my
main light source, so I can just go and
start moving it around. You can see if we're losing
some of that gradient. Maybe will bring it more say upfront or
actually more behind. So now the gradient there
is much less visible. Let me lighten up here. One
26. Leveraging Blender 4-3 Light Linking: Seeing right now, I'm making
the shadow too bright. To help with that, we need to
start using light linking. This is a new feature
of 4.3, blend of 4.3. I just need to go into
object settings and go into shading and into light
inking and click New. And then I want
to drag my plane, my shadow catcher inside,
and just unclick it. My plane is no longer
affected by this light. I can also do the same thing for this for the point light
that is hitting the face, I will go into shading inking, create new and dragon drop
it play again and click it. So now our shadow is
much, much darker. So it's a little bit of trick. But a lot of this
trickery is very, very common in
production where people hide stuff in every
possible way they can. Another
27. Adjusting Global Environment Colors: Shadows is to change
the global environment. So I want to go into the world and switch from gray to
slightly darker color. It can go if I go full black. We're no longer receiving any global illumination coming
from the gray background. And if I make it
white, everything starts to light up too much. I also want to make
this, you know, color a little bit more brownish to match up with the
background of the room. It's about right. Make
it a bit darker. I
28. Adding Bloom Effects in Blender: Bit of bloom to the scene, and in wonderful points
really changed the bloom from being very simple to a new a
bit more complicated setup. So I'm going back
into compositing tab. I don't need well, I need my view, but I don't need my map range or depth channel. I'll press pace but and I'll
look for the glare node. And by the way, I
have this render in the background because
I've just press render at the top and
introduce this render. Then I will go and connect
the image there to image and then to the view node and then
to final composite node. Then I want to just go and
drag from the left and split the screen and go click at the corner
through your view pods, and just do a little
preview of the bloom. So right now we don't
have any bloom preview, but if I click right
hand in shading, I can go and say always. So that composite will start
to activate inside my scene. So whatever I do here, it will be visible in the scene. Then I can start to
go first of all, I need to change this
trick type to bloom, and then I can start
to play these options. So I can just drop down the threshold and starts
to lighten everything up. What I like about it that
we actually also get in our AI render to
give some bloom, and that adds up
a bit of that td dimensionality and a bit of immersiveness,
the whole render. Then can go and
decrease or increase. This is up to you how strong you want to make it. An
29. Adjusting Brightness and Contrast of Model Texture: Tweak I want to
add, I will go into shading tab and look for
brightness and contrast. Can just type in contrast,
position in between, and then increase
the brightness and decrease contrast or
maybe the vice versa. So I want to have less of a contrast to achieve a
softer look of the model. And if I press on the node, it will mute it, and I can see the difference before and after. So it just adds a
bit of that washed out coloring as per
the background.
30. Animating the Camera for Dynamic Shots: Animation, I'll click
to show the overlays, click on my camera, and I
do want to split my screen, drag down that part, and I'll press nu zero to
jump to my camera preview, and I want to lock my
camera lock to view. Then I want to activate
a live keyframe. So every time I move
the camera now, it will create a
keyframe for it. Press G, move it and
it creates a keyframe. So then I will go jump to
say keyframe 60 and move our camera right to the
face of this character. And I can rotate it a
little bit, as well. So if you didn't key
the rotation before, that's a little bit annoying. What I want to do is right click inside the keyframe
to key the rotation, then jump back to
the 60s position, and I can just press shift and drag this to the left
and change it like that. I think I might be interested in changing the focal lens as
well. Let's play with that. So I will introduce
a keyframe just by clicking over there
and go to frame 60, click here, and maybe say, I'll go into mob portrait lens onto Zoom in right on
this guy, right on his
31. Adding Professional Look with Depth of Field: Mode. Usually, what's
going to happen, you'll get a blurred background. So I want to introduce
depth of field, and I will focus on object. So I'll just pick
this guy and we can play with the
amount of F Stop. It's not really
focusing that well, so I can just drop my object and only do the
focusing distance. So I can drop it down to zero. One press shift and trying to get that
focus. There we go. If I zoom back in,
everything is blurred out. So I want to also keyframe
the focus distance. Let's put it here to
something a bit more normal. Like say 2 meters right, and let's just play
this animation. I will limit it to 60 frames because we only have
a short fly through. It might be a little
bit too fast. I just need to pick that
keyframe and drag it all the way to 250 if I want to
make it to run slower. Forgotten to activate
composited view. So now it's much brighter image. Let's see if it's
not too bright. Looks a little bit like we have almost too much glare here. So we'll go into compositing
and I can play with the threshhold or size
of this blurriness. Bloom. Let's hope this is better. That's
pretty good. I think
32. Final Light Adjustments: Much of the light in
front of the face. So I'll pick the light
responsible for that brightness, and I'll just go and
drop it a bit down. It's a bit more
natural lighting now. To fill this window, I have to kind of go
and push it up to join the areas that we have
full screen preview. L
33. Rendering the Final Animation: A couple of monitors and
have it on my second screen. It's much easier to deal with
that on multiple screens. Now let's talk about the output. We can output as a PNG sequence. But for the sake of
time, I'll just go with mpeg expot as a video, and here you just need to pick
the folder to export this. I got my folder, and then I will go and say
render animation. Make sure you save
before rendering.
34. Final Animations Overview: Our full animation. There is another variation. Now it's an Egyptian themed room with the prop I've
downloaded from Sketchfab, or a different time period, in this case, using another
simple prop in the middle. So thank you guys for watching. I hope you enjoy this course and learn something new here.
Anton Tenitsky, 3D Artist
