Hey Guys,
I’m currently trying to add additive animations to my character. I’ve two animations:
So from my understanding, as soon I set the weight of the additive layer to 1 the character should play the idle with the hold weapon animation, but instead my character still idles without holding the weapon.
Is there any special trick to create additive animations?
I’m using Unity 5.3.
I’m not sure why exactly are you trying to use an “Additive” layer, I usually use an “Override” layer for that, have you tried it already?
Override is working fine, but it doesn’t look good when my character is running (the arms are too steady).
Hi silentneedle,
How did you created your animation clip?
Additive clip are kind special, if you import a clip to be used in a additive layer and doesn’t setup the additive reference pose, the system will use the pose at frame 0 in your clip as the reference pose.
There is two way to setup the additive reference pose, when you import your clip you can define a frame to be used as the reference pose
or by script with
I guess in your case you have an animation that hold a gun that last 1 frame, so the system use the same frame as the reference frame and the animation to add, since there is no difference between frame 1 and 1 you don’t see anything move
Hi Mecanim-Dev =D
Why is the Additive Reference Pose Clip setting not easily accessible for clips created within Unity?
If I hadn’t found this thread I would have continued to think the additive feature in Mechanim was broken.
For those struggling with this you can additionally set your editor to debug mode to expose the settings when you have an animation clip selected. Don’t forget to set the Has Additive Reference Pose bool as well as setting the Additive Reference Pose Clip otherwise it wont work.
Trying to do exactly this; I have an animation (let’s call it “happy”, frames 40 to 64) that differs significantly from idle, both clips from a single asset, and have the pose frame for happy set to the first idle frame (and additive reference pose flagged on obviously).
Unity would have to take happy and substract that idle pose to get the additive difference, right?
And if if I then play happy as additive onto idle, I should get my full happy animation, right?
But it doesn’t; it seems to just take the first frame of happy as a reference pose, and creates an additive clip based purely on that pose (which is some minimal jitter currently) that gets added onto my idle.
I would say the result is quite sad instead…
Am I missing something here?
Actually, exerions tip for using debug mode revealed the issue; regardless of what frame is set as a reference pose, the importer will make use of a ‘preview’ clip based on the clip used as additive animation. Any frame lookup will be done relative to this preview clip. Unfortunately, these additive reference pose clips cannot be set in the animation clips of an asset. The only way to work around this bug is to duplicate these clips to stand-alone assets, and then set the reference pose clip to an idle animation (or whatever animation is to serve as a basis). The downside here is that updating the original asset will not update these duplicate clips.
lol lose 2 hours before finding this thread. Additional clarification since this came up the top from Google, important rule :
Without entering Debug mode, if we have an additive layer with full weight, both base layer and additive layer playing single state
Situation 1
Base layer : single keyframe, looping, position x = 5
Additive layer : single keyframe, looping, position x = 1
Expected : constant position x = 6
Result : constant position x = 5, because the additive layer reference the first frame of itself, which has position x = 1. When it is playing, it is always 1 - 1 = 0 so nothing got added.
Situation 2
Base layer : single keyframe, looping, position x = 5
Additive layer : 2 keyframes, looping, position linearly ramp up from x = 0 to x = 1
Result : position ramp up from x = 5 to x = 6. Additive layer use the first frame of itself as a reference like before but this time it’s x = 0. Overtime we get +0 ~ +1 to the base layer correctly. You can also adjust the total addition with Additive layer’s weight.
Situation 3
I want to upgrade the ramp, so I made the 2 keyframes that was ramping up from x = 0 to x = 1 to become x = 3 to x = 4 instead.
Expected : position ramp up from x = 8 to x = 9
Result : position still ramp up from x = 5 to x = 6. Again, since the additive clip use the first frame of itself as a reference, we get 3 - 3 = 0 ~ 4 -3 = 1 → still ramping up from +0 ~ +1
How to make the first situation producing constant x = 6, and the 3rd situation producing ramping up x = 8 to x = 9?
Make a custom reference clip. It doesn’t have to be in the graph, but you should copy from your previous clip so it has a Position keyframe. (If it is not in a graph, you cannot key a new kind of keyframe to it from “nothing”, but you could delete keyframes) Make it a single keyframe with 0 0 0 position.
Enter debug mode of the additive clip with a single keyframe position x = 1. Connect the zero reference clip and check has additive reference pose check box. Now x = 1 will be 1 - 0 = 1 correctly and not 1 - 1 = 0, and produce a constant x = 5 + 1 = 6 when playing additively.
(imo this box should be exposed, especially that I am doing UI animation where there is a lot of single frame animation that I want them added it is not obvious that I have to enter Debug mode.)
I agree that this should be exposed. And thank you so much for this post! I’ve been trying to get additive blending to work properly and couldn’t until I got here. I found a lot of “use override” posts, which really don’t do the trick.
Thanks again!
Same here - this has been an issue in the past and the solution was usually awkward. Very happy to see an updated thread and have stumbled on this 
Hard to believe they couldn’t fathom that people would import/create pose clips for additive poses. Not hard to believe that they would neglect fixing it for this long. I hope people are happy with their Machine Learning and DOTS toys, because the tools are rotting.
Thanks for the workaround.
Sorry to bring this up again, but it ties into this and others may have the same thought/issue…
What happens if my animation on an additive layer is a loop that doesn’t ‘overlap’ the reference animation position/animation? If I ‘NEED’ an additive reference pose frame (say, frame one) but my character needs to be positioned away from the position of the base animation and is a loop…how do I create a reference pose for a loop without having a single frame look like it’s glitching because I HAVE to have a reference frame so it knows what to blend from?
I hope that makes sense?
I know this is very late but to do this I:
And it was job done.
It’s an absolutely garbage process though, and it’s embarrassing to see that it’s been like this for half a decade!
Yeah that’s the annoying part: you have to make a copy of the clip (ctrl+D), so it becomes a stand-alone asset. Only then can you assign it a reference pose clip.
Is this still the way to do additive layer animations in 2022? Having to go into the debug inspector is so clunky.
This is absolutely nuts. It’s 2023, why can’t we just drag another FBX into here to set this up.
It’s just the Unity’s way. Double down on the new shiny buzzword (yeah, they announced Unity AI, whatever that means.) and let the engine die.
I just found my new least favorite feature of Unity! I’m sure second and third place won’t mind since they’re likely animation related as well. The intended workflow is unintuitive, buried, and unexplained. I would never have found it or known it existed if not for this post.
Here’s my best summary of what I learned today. I’m not an animator, so if this isn’t quite right feel free to correct it.
Additive is presented explicitly as an interaction between animation layers, yet it appears to sabotage itself before it gets a chance to apply that totally normal behavior. Rest assured, it will add correctly… but only after you manually assign a reference pose within the context of the layer (unless you happen to like the weird default?). I personally expected the default to just be zeroes for everything, because that would make sense and allow me to proceed without the extra step. That’s not the case. There is a zero involved, but not the one you’re hoping for.
The default reference pose is the first frame of each animation clip (time = zero). If, like me, your animation clip doesn’t begin with or pass through a base pose, or if it’s a single frame state relying on transitions to blend, this default unusable. You could restructure your whole set of animation clips, or you could assign a pose manually.
Your custom reference pose should instead be a single frame animation clip capturing the values your clips relate to. If you’re like me, you might include every layer that gets touched and set them all to zero. Unity makes no effort to explain how this process is handled, or even that it should be handled. Keep in mind that this pose is relative to the current layer only, so you don’t necessarily need to pack it with an entire T-pose or all of your settings across the entire animator, though it might be easier.
The advantage of the custom reference pose is that you only need to maintain it once across all animations in the layer and you can change it later on to increment default values while preserving the offsets in the animation clips. I assume the default pose must be maintained across every layer. I’m sure there are plenty of animation styles where that behavior is fine and dandy, but we wouldn’t be here if we were using them.
To assign the reference pose, you need to select all animation clips used in the layer and set the inspector to Debug mode. Unity made a dumb dumb and buried core features where you’ll never ever find them and didn’t leave so much as a ransom note to let you know where to dig. Assign the pose to Additive Reference Pose Clip, then enable the Has Additive Reference Pose toggle. Now your animations will use the values contained in the custom reference pose to finalize the layer’s internal values.
When it comes to actually adding between layers, as the original workflow implies, the principles appear to be what you’d expect. The internal value (animation value - reference value) is added to the external value, which is derived from whatever layers are beneath the additive one. If none of those layers touch the same property, I assume you’ll have the same problem you face with override layers when a property is only sometimes assigned.
** Slight differences for imported animation clips versus standalone files. Looks to require copying to standalone clips.
** Not sure how an empty animation state behaves since it lacks a clip to assign poses to.
Things Unity should go ahead and fix several years ago:
And as far as I know, there is no way to set the reference pose at runtime since AnimationUtility is an editor time stuff.
Anyway bad design is the norm for Unity, so that’s expected.
