Unreal Engine Livestream - Character Animation in UE4 - Jan 25 - Live from Epic HQ

[DizzyWes]

Hey all,

Here is a follow up to the questions that were asked that I didn’t get to on the stream. Hopefully these will answer your questions, if not we’ll do a more advanced verison of this stream in the future:

---

Q: I want to layer/blend a melee attack on top of a walk. The problem is that I can only sample from the spine 1 up but not the pelvis root. This makes the melee attack look funny since it doesn’t have the “weight” behind the motion (slight pelvis up and down movement). What I want to achieve is 100% feet from the movement animation, 100% upper body from the attack, and like a 50/50 split on the pelvis. Can this be done?

Q: When you layered the attack on top of the walk, the pelvis didn’t follow the attack. It makes the attack look weird. How do you include the pelvis at 50% of the attack and 50% of the walk?

Q: Can you talk about the “Blend Depth” setting from the “Layered blend per bone” node.

WB: Hopefully answering all three questions with this response. The Blend Depth setting on the Layered Blend Per Bone node determines how far down the chain of bones you want to start distributing the blend from.

For example, if we use the following Blend Depth on our Layered Blend Per Bone node:

file:///C:%5CUsers%5CWES~1.BUN%5CAppData%5CLocal%5CTemp%5Cmsohtmlclip1%5C01%5Cclip_image001.png

It will take the 4[SUP]th[/SUP] bone in the chain and distribute the blend among the bones back up the chain.

The blend is stronger starting at the Neck, then down to Spine 2 and Spine 1, weaker when reaching the Spine bone which is why there is less twist in the animation with a higher blend setting (and more twist with a lower setting as 100% of the blend is happening at the Bone Name we specify).

https://www.youtube.com/watch?v=ap7r0PTcHXY

While you can’t do a 50/50 split of the Hips in this case since that is the root bone, there is a way to negate the effects of the Layered Blend per Bone node by entering a negative Blend Depth specifying which Bone or chain of Bones to ignore.

For example, if we want the hips (pelvis) to follow the attack, we can take both legs and add them to the Branch Filters, setting their Blend Depth value to -1. It will negate the Blend on the chain of bones starting from the Bone Name specified.

In this example, we add the Bones for the legs and set the Blend Depth to -1 so that the hips follow the attack better. You will have to experiment with different settings to achieve the effect you want, but it’s already an improvement over the previous setup.

https://www.youtube.com/watch?v=CGPez5NlT5I

Q: What is the performance impact of using Cached Poses in an Animation Blueprint? If we are caching poses to be used throughout our Anim Blueprint for various blends, is there a more efficient way of structuring our Anim Blueprints?

WB: The performance impact of cached poses is negligible compared to additional State Machines. Having to evaluate multiple State Machines is more expensive than taking a reference pose from the evaluation that already occurred on a State Machine.

Q: Could you provide input about the pros and cons of hierarchic states machines? (If you could explain them overall for the rest of the people would be sweet)

WB: Pro is that you can visualize better, you could see the flow of the State Machine better. More State Machines, however, mean more depth and more evaluation happening.

Q: Why not add a Montage to the animation state machine?? The whole purpose of the animation state machine is to play animations. Why play animations someplace else?

WB: Montages cannot be placed in AnimGraphs, I probably could have told the Character Blueprint to call an event inside the Event Graph of the Anim Blueprint to play the montage instead. You could, of course, skip using montages all together and use the State Machine to handle combat animations; it’s a matter of preference. In the stream, I discussed why it may be advantageous to use Montages instead of a State Machine and using the State Machine to handle systemic animation logic.

Q: Is an animation montage what you would use for adding facial animation during dialogue?

WB: You could layer that on top of an existing animation. You may also want to look at using Pose Assets to handle facial animation.

Q: If we were to have multiple animations related to multidirectional input (for example, slipping when changing direction) and the character also has impaired movement due to status or damage (covering a hurt arm, covering face, broken or hurt leg), those are a lot of animations or chunks of animations (like when jumping). It’s a lot to handle, should we do it through a lot of montages/blending or is there a “best practice” to handle a large number of conditional small animations?

WB: If they all affect the multi-directional navigation, for example, a broke leg and the character limps or hurt arm so the character holds it differently. I’d probably look at using blend spaces for the navigation and layer on top the upper body or lower body the conditional animations. You could also set up blend spaces for your conditions and selectively choose one of the blend spaces based on those conditions. Maybe we’ll cover this in a future stream as well, I’ll need to dig into it more.

Q: What would be the best way to handle Anim Montages in a multiplayer scenario? Should it be handled via RepNotify or simply just replicated from a Multicast in a Run from Server? What would be the best method for clean gameplay?

WB: I’d say the later, you could use a Run on Server Event to call a Multicast Event to play the Anim Montage.

Q: I hope he takes up the animation retargeting system cuz it won’t retarget correctly in 4.18… did they change anything for the latest update?

WB: Would you happen to have screens/video of the issue. Or a test case I could follow to reproduce it on my end to see what the problem is?

Q: I’d like to see something about the Control Rig and Mixamo Characters. Control Rig totally breaks down and warps my Mixamo characters

WB: Similar to the previous question, would you be able to provide some test assets for us so that we may diagnose the issue on our end?

Q: Can we maybe get a video tutorial on procedural animation sometime in the future?

WB: We can look into this. If it is not a video tutorial series, perhaps we can cover it in a future stream. I’d like to get Laurent to do one so I can learn from him as well! J

Q: What is the right way of implementing character roll/prone animation, considering root (capsule for example) component? How should mentioned animations handle with their parent capsule on a character?

WB: I personally do not have a best practice for this just yet, however, I’m hoping that in future releases when we adopt some of the techniques being used in Paragon that we’ll have better features to handle this.

Take a look at the video I talked about from Laurent here at the 12:00 mark where he talks about Movement Prediction, Distance Curve (reverse Root Motion) and Speed Warping. I don’t have a timeline of when these features will be available, however.