Week Four of Genius Hour Two - An Early Valentine's Day Special

This week, I modeled a rose!  This was in order to practice difficult modeling techniques and so I would be able to make plants for future projects.  I used a lot of reference images for this rose but chances are it still looks a bit fake.  It was good I decided to do this instead of another scene.  There was not enough time for me this week to do a large scene.

To start, a cylinder was made and scaled appropriately.  The thorns were really easy.  A couple of edge loops were added to divide off a section.  Then, an edge within this section is deleted and a face made.  This gave me a bit thicker thorn than what a single face would have done.  This was then extruded out.  Similar to my parmesan cheese shaker last week, I needed to extrude the face, hit escape, and then extrude again so the surrounding mesh wasn't distorted.  A face is scaled in so a point is created.  Finally, an edge loop allows me to add a curve.  As I went up the rose, I tried to make the thorns smaller.  After the fact, I realized I probably could have used knife cuts instead of edge loops.  This would have cut down on the poly count.

I had this weird issue with the subsurface modifier not smoothing out an area like it should.  When turned off, you could see this shading issue.  At first, I tried to delete the face and make a new one, but the problem persisted.  Eventually, dissolving the face worked.  I don't know how it happened, but a face probably got doubled up.

The Bezier circle was probably the most complicated part of this.  I got the idea to use it for petals from this hair tutorial which I referenced before.  The video shows me messing around with the circle.  Bezier circles are used when you want a shape extruded along a path.  There are three different paths, Bezier, Nurbs, and path.  I choose a generic path because it worked the best.

After the circle is reshaped and extruded, I began to modify the path.  By scaling different parts and moving other parts around, I could get a petal looking thing.

A little curve was added to the petal.  It was also rotated before being applied.

The first few petals were small and placed in the center.  As I went out, the petals became larger.  Some petals were purposely made smaller than others to give it a natural feel.  Each petal was simply a resized duplicate of the original.

Time for materials!  In the left photo, I was messing around with shaders.  You can see a toon shader being used.  I plan on using it for its true purpose in my next blog, so be sure to check it out.

The rose was looking a little hollow, so I cut an isohedron in half and put it inside the rose.  I gave it a darker color than the surrounding petals.

To color the thorns, I had to use the same technique I used in my last project.  To begin, I gave a material to the entire stalk.  Then, I selected all of the thorns in edit mode and assigned a new material to them.  I expected some problems with the coloring, which you can see.

I wanted the rose to be in a basic scene, so I made one.  I had found a free texture website, so I decided to use one of the textures.  It came with a normal map as well.  This was a little confusing to set up, but I managed to do it.  The texture was tileable.  This means there are no visible lines or breaks when the image is repeated.  However, the mirror kind of messes it up, so trim is added to cover the bottom.

The pillars are simply one-fourth of a cylinder.  I had this distortion occur for some reason.  When I raised the subsurface modifier, the number of checkers increased.

Unfortunately, the image texture got all messed up the next day.  I really have no idea what caused this.  It may be a little hard to see, but the normal map is not distorted like the image texture.  This is why you can still see a ghost of the tiles.  Finally, I changed this floor to a checkered pattern and added a few other basic elements to the scene.

Kyle suggested a glass cloche like in Beauty and the Beast, so I did just that.  I also wanted an infinite mirror effect, so I also did that.  At first, It wasn't working.  This was because I lowered the max number of light bounces to make previewing a scene easier on the computer.  However, this meant the light was not reflecting off the mirror and onto another mirror and so on.  As such, despite the scene being relatively low-poly, it still took about three and a half hours to render because of the high number of light bounces.

I decided to play around with the camera settings for this scene.  Instead of the normal perspective camera, or basically how you see, I choose the panorama camera.  This allowed me to achieve that warped look.  As I said before, the camera in Blender acts a lot like a real camera, and there are actually preset settings for various cameras.  I probably won't mess around with them too much.

Like all the other scenes, I ended up rerendering it.  The glass was reflecting too much light and being really ugly.  To fix it, I changed its IOR.   I also felt as though the rose stalk and leaves were too green.  It looked fine in the preview, but the light seems to be bumping the brightness up.  Unlike the other scenes, I didn't feel like I needed to change the sampling to remove noise.  

Since I haven't explained it yet, I go over it now.  When an image is rendered, the computer renders it in sections.  The size and order of these sections is up to the person using the program.  I set the size to 256 by 256 as that is the best size for GPU rendering.  What determines how many of these sections are rendered at a time is the number of threads your computer has.  My dad's Mac has eight, so eight 256 by 256 squares are rendered at once.  The sampling determines how many times the computer goes over that section to achieve an accurate image.  This means that a higher sampling number will result in an image with less noise.  That also means, on the flip side, that it will take longer to render.  Kyle knows more about computers for sure, so direct any questions about GPUs and CPUs and threads and whatnot to him.

You may have noticed that some of my screenshots contain parts of the rendered scene.  This is just a preview, so it is not that accurate and can have a lot of noise.  This is why I had to change the glass.  It looked fine in the preview, but not in the final render.  I can also render the image at a smaller size with a lower sampling size, but this is often very noisy and is usually worse than the preview.

Here's the final image!  And here's the Google Drive link to download it.  The glass looks much better, but I probably should have lowered the IOR even more.  I noticed on Blender Artists that many people said they ran their images through the denoiser.  I assumed this was some outside program, so I mostly ignored it.  Some research into the subject proved me wrong.  The denoiser is a part of the Cycles rendering engine and can be turned on with a single click.  I kept wondering why some of the art people would post would be so crisp when they said they only used a thousand samples or less.  The denoiser seems to average the color for an area.  As such, it appears smoother.  On the other hand, it adds a weird look you might not have noticed at first.  The pillar is one color and there should be enough light to make appear as such, but the denoiser has sort of caused a swirl pattern.  It is hard to notice, but it definitely looks better than a bunch of noise.  I will keep using it, especially because it doesn't affect render times all that much.

I know it is about a week early, but here is a Valentine's Day card!  Download it and send to your girlfriend!

I mentioned Tigtone in my first blog.  Unfortunately, someone deleted the recording so you will have to wait until the next blog before learning about it.  It is very inspiring to know that garbage looking CGI like that can be passed off a show.  I probably can't do much better, though.  I wanted to add an image, but it was acting kind of weird, so simply Google it and look at some videos. 

Speaking of CGI animation, I plan on doing some next week.  I will make a character, color the character, and then have it do something.  It will probably be a robot.  That way, I can keep it low-poly to save on rendering time.  It is also because I don't want to spend hours trying to make a person and have it turn out poorly.  I should probably learn how to draw people in 2D before I make one in 3D.  

I will try to use cel-shading on the model.  This is a technique used to give CGI animation a flat, 2D look.  It is used in movies and games.  In movies, it is used to easily animate complex scenes and camera angles while tricking the viewer into believing the scene is still 2D.  In games, it gives it a stylistic look.  Probably the most notable instance of it in video games is in The Legend of Zelda: The Wind Waker.  I always liked its art style when I played it, and thankfully did many others.  Therefore, I won't have too much trouble doing it.  There are many different ways from what I have seen.  On the other hand, this may mean that I will have to fiddle around with each way until I get the desired result. 

I am not really following my action planner.  I should have been finishing my donut scene this week.  I gave myself a lot of time to do it just in case.  After that donut scene was supposed to be Blender Guru's anvil tutorial.  It shows some advanced modeling techniques and goes over how to properly UV unwrap a mesh and bake a normal map.  I simply wanted to move on to my own scenes instead of copying another tutorial.  I also didn't want to watch another couple hours worth of tutorials.  However, it might still be worth my time to watch it in order to learn how to bake my own normal maps and to learn how to properly apply an image texture.

Either way, tune in next week for a CGI robot!