Part1 Key Animation
This week I did an animation, which is the action of flower arrangement. I found flower models on the Internet, because they are all merged together, so I separate them one by one to facilitate the production of animation.
For this scene, I also added two facets to make the character more prominent.
I think the difficulty of this animation is the parent relationship of the object. Because the process of this animation is that the character picks up the flower, and then the flower will be put in the vase. During the period from holding the flower in hand to putting it down, the flower is always on the controller of the hand by parent relationship.
My approach is to confirm the animation, copy a flower model at the frame where the perianth is picked up, and then key the visible and invisible attributes. Then the flowers that are not bound to the parent-child relationship are not visible at that frame, and the flowers that are given the parent-child relationship are visible at that frame. In the same way, at the frame when the flower is ready to leave, a still model is copied, and it is visible at that frame.
I have to admit that this is a very stupid method, but I didn’t think too much at that time, because the method is really feasible. The only drawback is that once I want to adjust the arm movement, I need to adjust a lot of things. I have to redo the basic flower position and keyframe. Later, in the course of class, my classmates raised this question, and I found their method better and more convenient. I think I may use this method later, because it is more scientific.
By the way at the beginning, the action was about 19 seconds, but I think as an animation played in the game, it should be short and general, so I accelerated a little bit.
Playblast — rose
Render –rose
Part2 Multiple Mocap Retarget
Based on our decision to set touch play animation in the game, we discussed it with the team members again. We feel that we need to make some adjustments to the reference settings of the previous shooting. One reason is that we think the previous reference length is too long, and the action is relatively simple, so it will be a little boring. So we shortened the length of the animation, and I proposed that we can use some motion capture animation, because I have spent a little time to understand this before, and I think now is the time to make use of this knowledge.
We screened some actions from “mixamo ” together, and then what I’m going to do this week is match the bones of the bound character with those of the capture character. (https://www.mixamo.com/#/?limit=96&page=1&type=Character)
First of all, I changed the role that was well bound and debugged without many problems to T-post. Then I started the task of matching the skeleton to the control panel on the right.
In this process, I encountered a problem, that is, all the models I have contacted before can be automatically recognized symmetrically, but I don’t know why, even if I confirm the naming in the outline view.
I asked Luke, and he told me that it was because the bones were not symmetrically bound, but the binding had been completed by this time, so I had no way to modify it.
So I debugged almost 10 times, but still failed. But I don’t want to give up this, because as long as the problem of this arm is solved, other problems are basically very small.
So I tried several ways later.
- I first matched the bones of the right arm, then matched the left arm, then made clear the left arm, and matched the left arm separately.
- Then I removed the shoulder bone.
- Then I removed the upper arm bone.
- I’ve even tried to see the remodelling from the chest.
In a word, I tried for 3 nights, about 10 hours, but I still have no idea.
Later, after I tried for a few nights, I suddenly realized whether it was because I chose the bone in the view. In fact, the standard practice should be to select the corresponding name in the outline view, because if you select it in the view, it will select other bone points, but I can’t notice. It’s really strange that there are two displays of this bone in the outline view, and the correct one is to find the right bone layer by layer.
On the other hand, I spent a lot of time matching my fingers.
Because I’m not sure how to match. So I drag the downloaded model into Maya to check, because since I want to match the two, the bones of both of them are consistent. So after checking one by one, I found that I still had to choose the standard bone name in the outline view.
reference–character2
character1 (custom rig)
It can be seen from this that I have really tried many times. At first, I will save it. Later, when I failed, I just kept rematching without success.
It can be seen that there is still a problem with the binding of my arm. Even if I brush the skin and weight again, the problem has not been solved. Later, I thought that maybe I didn’t choose the right controller, so I referred to the binding of other characters and our own characters, which should be the shoulder problem. So I cleared the previous match and redone it several times.
Then I matched the controller again. At the beginning, the action was very strange. I thought my controller was wrong. I went to watch the video again, but it didn’t seem to be correct, so I didn’t use the tutorial method. I canceled the knee match and checked my shoulder binding more carefully. The revised one is really much more comfortable.
The final matching effect
After changing the position of the controller, there are obvious differences between the two versions. It seems that the arm before the change is more uncontrolled, and the one behind is much better.
Comparison of the two versions
Mocap animation
It can be seen from the above animation that most of the animations have been successfully copied to our model, that is, we need to adjust some key frames carefully to make the action more comfortable.
Part3 Conclusion
Another difficulty in this animation is that my rigging is not very good, so the posture of my arm is always uncomfortable when I lift it up. To solve this problem, I spent a lot of time adjusting the skin and weight, although the effect is still not obvious. In addition, she didn’t have a detailed expression rigging, so I cannot do eye closing and mouth movement on her.
I hope that if I have time later, I can take the time to modify the character, make the facial expression better and make the skin more accurate.
At last, I was very happy to be able to do this Mocap successfully, because I felt very anxious and irritable when I didn’t solve the problem in those three nights, but fortunately I didn’t give up. After that, I will copy some motion capture animations we found to our models. Anyway, this step saves a lot of time and energy for our later work, and makes me more confident in completing our project.