Storyboard and character action production:

This week, Karl, Sean and I confirmed their respective progress. Before, I spent a lot of time on asset production and workflow research. Now I can basically start the production of specific content.

The first is to draw a dynamic storyboard to determine the basic atmosphere feeling.

After discussing it, we all agreed that music is a very important part of this kind of short films that are not very plot-intensive. I contacted a younger student of mine who studied music in London before, and after describing our plot to him, he shared a piece of music that fits the theme very well. The arrangement and structure of the music are also very suitable for the rhythm of the short film.

Based on this music, I made a storyboard in ps, and added notes in PR.

This is the final storyboard video:

Regarding character animation production, I decided to use the mocap animation data redirection production method. First, I listed a list of actions that may appear in the movie.

Then I searched for suitable actions in mixamo, then used HumanIK redirection in Maya, and added the modification details through the animation layer. Take the Idle state as an example. To check the action pose, I uploaded the model to the Mixamo website and adjusted the parameters to get a basic animation.

Choose withoutskin when downloading, so that you can get only the bone data.

In Maya, HumanIK first creates a character definition to define bone information.

Then create custom rig, select the controller created by advskeleton5.

Then import the downloaded mixamo data and create a new character definition.

The character gets an animation! Then bake animation to a new animation layer.

Then I created a new animation layer, modified and adjusted the details in it, and added the statue held by the character. This is the final animation effect:

After importing it into the UE4 engine, combining with the scene made by Karl, we got the first shot:

This week I did a test of some experimental functions, mainly the realization of the effect of three-sharp two in UE4 and the binding of the muscle system in Maya.

Three-rendering two is a very popular technology at present. By changing the rendering style, the same model can produce some stylized effects. Here I mainly learned the methods used in this video.

It is a kind of post-processing effect to make the model look two-dimensional. The basic idea is to judge whether the colour of the current area is under a Threshold value and assign the same value to the adjacent brightness. Through this idea, the light and dark areas of the model can be clearly divided. Setting a few thresholds will have several layers of shading effects.

This is a threshold setting method.

Generally speaking, there are three layers of light and shadow in Japanese anime. Here I set eight layers according to the value of the video.

The final effect is okay. But the dividing line between light and shadow has some graininess, not smooth lines. This is a problem that I need to improve in the future.

Regarding Maya’s muscular system, it is a technical point that I considered when binding. Maya has its own muscle tool.

The basic logic is similar to bone binding, but the difference is that each part of the bone binding is controlled by bone points, which is convenient but difficult to achieve better results. For example, when the arm is bent, the inner side of the elbow will be compressed and deformed, which is more difficult to achieve under the bone binding.

The muscular system assigns the weights of various muscles to the model. An arm is no longer just controlled by two bones but has corresponding changes in different areas.

I mainly tried to create the biceps of the arm.

It can be seen that the skin of the arm is supported by the muscles. But the main reason I didn’t continue to finish was that the skin drawing of the muscle system was too stuck to run on my laptop, and it was time for a new computer.

In the end, I integrated all the things that can be shown in my personal projects in UE4 and used Sequence to create a video, similar to the process of creating shots in Maya. This is the final result video:

With the release of UE5 Early Access, Unreal Engine has a new amazing performance in graphics and production. I have communicated with some people in the film and television animation industry and learned that in terms of the current animation industry and game industry, Unreal Engine technology is actually still a brand-new technology, especially in the design and production of 3D animation, there are still many The content is worth exploring and researching. Gradually applying Unreal Engine technology to 3D animation design is a major current trend. Combining my future goals for TA or TD related work, I need to further study Unreal Engine’s solutions in various types of work.

What I mainly read is the official documentation of Unreal Engine and the literature of other people using UE4 to make animations. In the official document, there is a fairly long description of animation production. Among them, there are many methods for character binding. Some use Maya to bind with plug-ins, and some use Control Rig to bind in UE. In addition, some physics-based animation and automatic IK animation are also mentioned. These documents provide a lot of ideas, but they are more basic, and I still need to explore the specific application links.

In addition to this special technical content, I will also study character design and animation design in some games. I will continue my personal project for the third term. I will mainly watch some Pixar and Sony films on the role performance. design. In terms of art style, I will mainly pay attention to some works of Alberto Mielgo and try to study his production process.

http://www.albertomielgo.com/

According to the development of my cooperation projects and personal projects in the third semester, I decided to focus on the research of animation production solutions in UE4 in FMP. The goal is to fully use the functions in UE4 to solve character rigging, animation production and rendering. Waiting for work, finally got a stylized character animation.

Animation Design：

I originally wanted to make an animation similar to Highlight Intros in Overwatch, and I mainly referred to this video:

But later with research, especially when I saw the analysis of Overwatch animation in this video,

I think that if you want to make a good animation, you should first start with the design of Pose. In the process of designing the character pose, there is a consideration of the character expression and background story of a character. Why does she make such actions and expressions like this? What state and so on.

In the end, I decided to start with designing the victory pose of the character.

Looking back at the character I made, and recalling my design ideas at the time, I wrote a simple character bio for this character:

She is a child of an engineer. She has been interested in mechanical engineering since she was a child, and she is very good at learning. In her spare time, she is an excellent roller skater. But in an accident, she lost her parents and her legs. She was adopted by her parents and colleagues to create a pair of mechanical legs and four powerful mechanical arms for herself. After experiencing many hardships, she still maintains an optimistic attitude. To save more people like herself, she joined an international organization to carry out a lot of rescue work.

After sorting out my thoughts, I think her pose should highlight her mechanical arms and mechanical legs, which are indispensable equipment in her rescue work. Her expression should be an optimistic and confident smile, making people in difficulties feel a positive smile. So this is the victory pose I designed after thinking about it.

This week I made the character hair, mainly using Xgen. During this time, I watched the video tutorials that Karl learned before. This is very helpful to my work. By the way, the work of UV and model optimization is in Karl helped me to complete it before. He is very proficient in the use of RizomUV. Thank you, Karl!

Hair making is not difficult, but it is a job that requires patience. The basic production step is to first extract the polygon face that will grow hair. On this character’s face, there are eyebrow, eyelash, hair and moustache. As well as his face mask and cotton hat.

The first is a cotton hat and face mask, because he grows hair in all positions, here is the creation of the groom in Xgen.

After creating, you can edit the hair density, length and width in the Xgen editor. Then only need to add a noise effect to the modifier, you can quickly make the fibre effect on the clothes.

In addition to the previous steps, it is necessary to add a mask drawing to the production of facial hair, because the hair has obvious density changes. Take the production of eyebrows as an example. If it is directly generated, hair will grow in all positions, which is obviously incorrect. Therefore, you need to click the brush icon to the right of the mask to draw an area, here is similar to the weight drawing, the whiter the place, the denser the hair.

The direction of the hair also needs to be adjusted. In Xgen, the guide is used to control the direction of the hair. After that, the specific trend of the hair needs to be controlled with a brush. In addition, there are other functions, such as mirroring guides or adding guides. In short, this is a very need for patience and Careful work, you must slowly adjust the hair trend to the ideal state. The time spent here is to achieve good results later.

Final Result:

The final overall hair effect:

The effect of Arnold rendering in Maya after adding materials:

The eyes are not made of material here because we decided to use the original inspiration source as the character’s eyes.

The next step is to import the hair into the UE.

Here is my main reference video

Maya export groom to UE4:

Groom import to UE4:

Hair Shader in UE4:

One of the main problems that need to be solved in UE is how to make the hair follow the animation. In the reference video, UE4.26 provides a very good binding function. I imitated the operation in the video. Most of the hair is normal, except for the eyebrows and eyelashes. When I added the binding, the hair immediately became very flat. But the hat and face mask look normal.

I finally found an article explaining this problem after I searched Google for a long time:

https://answers.unrealengine.com/questions/1018697/view.html?sort=oldest

Simply, this is also due to the accuracy problem caused by the problem of character physical asset settings, similar to the problem of cloth simulation. This is more common on short hair, so I modified the character’s physical assets to only keep the collision body of the head.

Cool! Now they look normal!

This week I mainly carried out the role import work and cloth settings in UE4.

Role import is a simple but error-prone step. As I mentioned in the previous blog, I always get an error when I import it into UE after binding, that is, the imported bone should have only one root node. I finally found that it is because the bones of the face are independent of the Root_M bones. , This is a problem caused by the use of advanced skeleton5 binding, in which you need to pay attention to check the game engine under the face rig menu.

The material configuration after importing is not complicated. Export the Unreal packed texture from Substance Painter, and you will get three basic textures, basecolor, Normal, OcclusionRoughnessMetallic, and models with light-emitting parts that will also have an emissive texture everywhere.

These textures are well understood to use according to the name. One thing that needs to be paid attention to is the file with the long name, OcclusionRoughnessMetallic. It looks like this:

It looks strange, but if we look at the export settings in Substance Painter, we will find that the three channels of this texture store different information, AO, Roughness, and Metallic.

Therefore, the texture in the UE’s material should be connected like this in the picture:

After successful import, I get such an effect:

Next, because the outermost apron of the character is made of cloth material, I decided to try the cloth simulation in UE. I mainly refer to this video:

The basic process is to select the model to create the cloth data file, and then use the brush and other tools to draw the weight of the cloth simulation. I mainly painted the hem of my clothes.

The result of applying cloth after drawing is not good. This is because of the problem of cloth collision body setting. The imported skeletal mesh will generate a default physical collision body, which is basically automatically generated according to each bone, with some size and position. deviation.

Default physical assets:

Physical assets after manual modification:

(The part of the physical assets that will not collide can be deleted, and the fewer collision bodies can provide higher collision accuracy)

After that, you need to adjust the parameters of the cloth simulation. Here I just increased the value and friction affected by the wind, and the final result will make the cloth look a little heavier.

This week I suspended the modelling of the character because the next step is about the hair production of the character. In the previous discussion with Karl, we decided to use Xgen to make it in Maya. It takes some time to learn. We have discussed and determined the feasibility of this production plan, because ue4 supports the import of groom, so we mainly refer to this video for the follow-up hair production process:

This week I mainly researched and produced the rain effect in UE4. This is the video I mainly refer to:

(This is the first video in the tutorial series)

According to the content of the video, I summarized the idea of making the rain effect.

The first is the moisture effect of water, which can be achieved by reducing the surface roughness of the object. But this only produces a reflective effect. In fact, in real life, after an object is wetted by water, if the water absorption is strong, the surface colour will deepen, so a Porousness attribute needs to be considered. In addition, objects in the rain, especially light rain, will not all get wet, so you also need to consider adding a mask to control the location of the water wet.

Material Functions are little snippets of Material graphs that can be saved in packages and reused across multiple Materials. Material Functions allow complex Material Graph networks to be saved out and quickly reused in other materials, while also allowing more complex networks to be abstracted into single nodes, making material creation easier for artists.

Best of all, edits to a single function are propagated throughout all networks which use it. So, if you need to make a fix or change to how a function works, you will not have to make further edits to the many materials which may be using that function.

In this material function, the lerp node is combined with the mask to control where the object is dry and where is wet. Linearly interpolates between A and B based on Alpha (100% of A when Alpha=0 and 100% of B when Alpha= 1) The main influence is roughness and specular. The base colour channel reduces the saturation and brightness through desaturation and combines the porousness property to control the colour depth of the wet position.

Next, you need to make water drops on the surface of the object. In light rain, some objects such as leaves and the surface of the table will accumulate some small drops of water.

It is not difficult to achieve the effect of water droplets. The effect of water droplets can be achieved by using a normal map and adjusting the roughness of the material.

But there are other problems. First, the water droplets should only exist on the top of the object, that is, the surface facing the z-axis (the upward axis in the UE is the z-axis). Drops of water in the vertical direction will slip off. In addition, the water droplets should have the effect of dynamically disappearing and reappearing, because the rain is constantly falling, and the water droplets will be constantly updated.

First of all, on the first question, in UE, you can make a z-axis mask by getting the world coordinates. In UE, RGB represents the XYZ axis. After knowing this, I only noticed the colour of the coordinate axis in UE. It corresponds exactly to RGB. The same is true in Maya, where the y-axis is the upward axis.

Regarding the second question, the animation effect can be achieved through the time node. Basically, the time node is needed to achieve dynamic effects in the material. Here, by creating a speed property to control the speed of the animation, combined with the information in the blue channel of the texture, the dynamic effect of the water drop can be offset, so that it is more realistic.

After finishing the effect on the horizontal plane, you also need to consider the water flow effect on the vertical plane. There should be a water flow effect on the x, -x, y, and -y axes. In fact, the basic idea is similar to that of making water drops. First, a normal map is needed to achieve the convex effect of the water. The blue channel is given a different black and white colour or offsets in time, and then the world coordinates are used to make the mask, and finally use the time. node. Although the idea is similar, the work of each part has become more complicated, the first is the production of the mask.

The water flow mask not only considers the time offset, but also pays attention to the fact that there should be no water flow on the horizontal plane, so it is necessary to make an extra mask to delete the mask area on the z-axis.

In addition, the effect of puddles is also made in the tutorial video, but because this time I mainly play characters, the effect of puddles is not needed.

Afterwards, a complete rain effect material function can be obtained by integrating the material function.

You can nest and use the material functions previously done to achieve complex effects.

In the end, in the actual material application, only a final material function Rain_Final_FUNCTION needs to be added. By creating parameter attributes, you can modify them in the instanced material.

In the end, I can get a good rain effect.

This week I mainly did modeling and texturing and rigging work.

This is my final result:

About modelling and materials, there are no special difficulties that need to be told, but the main thing is to make it carefully. In the process, I mainly use Maya, Zbrush and marvellous designers.

In this process, I also mastered some substance painter usage skills, such as smart mask.

Smart masks can have a more natural texture effect than hand-drawn masks through some calculation methods, such as noise or finding edges. In this way, I can more easily add rust and dirt effects to ordinary metal materials.

To make the character’s skin more textured, it is also necessary to make the character’s scattering map. Subsurface scattering is a complex phenomenon that occurs in a medium with a high scattering coefficient. Common subsurface scattering media include milk, wax, skin, etc. After light is incident on the subsurface scattering medium, more Sub-scattering, and finally exiting, forming a special translucent effect.

In sp, you can simply draw scattering black and white textures to provide the light penetration of the character’s skin. Of course, this texture cannot be used directly in ue4. Here I refer to the explanation of subsurface scattering in this article.

https://zhuanlan.zhihu.com/p/337384739

The whiter part indicates the better the light transmittance.

In UE4, the subsurface scattering model can be selected in the created material, and the subsurface colour channel can be used in the material after it is turned on.

In the official UE4 documentation, the description of this channel is:

The Subsurface Color input is only enabled when the Shading Model property is set to Subsurface. This input allows you to add colour to your Material to simulate shifts in colour when light passes through the surface. For instance, human characters might have a red subsurface colour on their skin to simulate blood beneath the surface.

Therefore, if only the black and white image is connected, the effect will be black and white, and there is no red light after light transmission. This light red needs to be added additionally.

Because this character is not an ordinary two-legged character, the binding is somewhat different from my previous operations. In the beginning, I planned to manually create the bones of the four robotic arms, but this is very difficult to control, and on the other hand, additional controllers need to be created, which is very time-consuming.

But I noticed that additional bones such as hands or tails can be added to advance skeleton5. I chose to add additional hands, but the result is not good.

It only creates a new hand, no arms, and is directly connected to the root node.

But unexpectedly, I thought of directly copying the existing arm bones. I didn’t expect advskeleton5 to support it! ! And it will automatically modify the name of the bone.

In this case, when I clicked build, he not only generated the bones perfectly symmetrically but also generated an excellent controller!!

IK and FK conversion are automatically created for each arm. Thanks advskeleton5!

Usually, I will leave advskeleton5 after this step to manually modify the weights. But this time I decided to further study its other functions. So I saw his excellent weight distribution function.

Create a skincage to control the weight distribution of each bone, and then copy the weight of skincage to the character. This function really gave me a deep understanding of the role of copy skin weight. The other functions are just adjustments to the details of skincage.

One simple step made my riging work several times faster. Thanks again advskeleton5! !

The final facial binding is only one step different from usual because you want to import it into UE4 at the end, so you need to check the game engine.

A painful lesson, a problem that was discovered only after the import failed many times.

The work has been done well this week, and the animation production and cartoon rendering experiments will be carried out later.

This week I mainly did character modelling work. After communicating with Karl, we decided to modify the existing resources, to quickly get a basic shape.

Character:

A man wearing a face mask, have moustache

Cotton hat

Jackets

Mountaineering backpack

Here are references

First, I drew a draft to determine the basic shape of the character:

Then I found some free assets on Sketchfab, and Karl also gave me some of the assets he found. This is how it looks.

It can be seen that the basic shape is ok, but it is still very rough. Some parts, especially jackets and pants, are difficult to modify directly through ZB or Maya. So I decided to remake the clothes in Marvelous Designer.

Thanks to the use of MD learned last semester, this time I can directly start production. Both pants and masks are made faster. But this is my first time making a double-layered jacket. This article has helped me a lot:

Part 02 | 3D Cloth Modeling Tutorial

The main difficulty in this is the topolygy of the clothes model and the collision between the backpack and the clothes. For Topo, I mainly used the reset grid that comes with MD. Although there are some flaws, they can be further modified in ZB. ZRemesher can bring better results.

Regarding model collision in MD, I refer to this video:

Particle distance controls the distance between each point during simulation. The smaller the value, the higher the accuracy and the reduction in penetration. But it will bring a huge amount of calculations. In the end, I only used 5, and it would be too slow to operate because it is too small.

You can see that the clothes on the shoulders and chest have some errors, but it doesn’t matter, I can delete the inner faces in Maya.

Karl and I also discussed the main technical points in UE, which is what we need to learn and make in the future.

Technical points:

The character and scene texture in the rain

The dynamic effect of rain on objects

Character hair

Character cloth simulation

Full-process animation creation based on UE4.

Regarding my personal project, I had some ideas in the second semester. When I was making the second semester’s cooperative project, I was considering using some of the assets at the time for the projects in the future. The project that always starts from 0 will spend a lot of time in the preliminary work, but my technical skills are difficult to improve (from the experience of my undergraduate).

My inspiration actually came from this video:

In this video, the speaker introduced the animation pipeline in Overwatch in detail. I am very interested in the process.

In fact, during this period, I gradually determined my future research and work direction. I will learn the technologies based on the real-time rendering engine, and with TA or TD as my career goal. Therefore, in this semester, I will create a personal project that is convenient for my future development. I will use a variety of DCC software to make assets, and finally render them in UE4, making shaders and materials in UE4. The final effect should be a series of animations similar to the play of the game animation or victory Pose animations in Overwatch.

In the first week, I mainly found some references and simply drew a concept picture of the character: