Этот сайт лучше всего просматривать в современном браузере с включённым JavaScript.

Auto place mesh vertex?

fantasyz

Am I the only one think this task could be automated?
Just auto placing the dots for a mesh outline is already a great help when compared to 4 dots at the corners.

It can be done in the old fashion way (algorithm) or AI.
Not to mention AI is so commonly available these days, training a bot to help auto place vertex should not be too difficult. Having enough training sample is another issue though.But I think Esoteric should have lots of rigged assets each with lots of meshes so I am not worry about that. Just need to write a utility to convert those assets into AI readable form. :clap:

Erika

Spine already allows the auto-generation of points in a mesh, you can read more about it here: Meshes - Spine User Guide: Generate
I personally still think an AI couldn't do better than what a human eye can really do in way less clicks though 😛 .

fantasyz

I know about Generate. It is not what I mean.

I am talking about the shape of the object. By using visual analysis and auto place vertex based on the object contour, it could give a relatively good outline and user can start fine tuning form there. It will save lots of time when compared to just a 4 vertices square box given.

Contour detection is quite robust these days. Just need a little clean up and points could be placed automatically around a part.

For making good looking animation, lots of parts need to be done in mesh for deformation. I found myself spending quite lots of time outlining parts when creating mesh. Putting points around a shape is simple but tedious.
IMO, this step is almost inevitable when dealing with mesh and once automated it could directly save lots of time for the user and will be one big feature when compare to other animating software like DragonBonePro.

Nate

It takes only a few clicks to make the outline of the mesh. Doing it automatically is almost never as good as doing it manually, especially when minimizing the number of vertices. If you don't care about how many mesh vertices are used, automatic would save time, but otherwise you'll want to do it manually even if automatic mesh hull detection was available. It is easy for a human to make decisions about placement while keeping in mind how it will deform and the tradeoffs when including more or less whitespace vs more or fewer vertices.

fantasyz

Its definitely more than a few click...for shape to be bend smoothly.
The objective here is not to be as good as the work done by pro but to have at least a rough shape to start with in order to save time.
About the number of vertex, it could just be a matter of sliding a slider to control the desired density on the lines.
For the automation, I am talking about automating the contour only, not the inside of the mesh. The inside is of course better be done by a human brain as who know how thing will deformed.

From my perspective, it feature has great potential and if one of such animating software implemented it, the others will suddenly definitely be less competitive because instead of other features like clipping, this step affect lots of users.
If dev think its not worth it, its ok. Just pretend I didn't ask. Haha

Nate

The thing is that the placement of the vertices on the hull is important for deformation, just like the internal vertices. In some cases it could be done automatically, but many (I'd even say most) cases are better done manually. Still, we'll likely implement the feature. I'm not arguing against it, just explaining why we haven't prioritized it higher. 🙂

RuinerXL

Sorry for necro-ing this, but I've seen this feature talked about a few times over the last few years and the response that it "only takes a few minutes to outline" is kind of poor. It may take a few minutes to do it once, but not 2000 times over the course of the full asset library for a game. That's thousands of dollars down the drain.

Considering this feature is now available in a lot of software that doesn't even make animation a priority (Photoshop even has this for its puppet warp), it just seems like an oversight that it's not in $300 animation software. We're using pose-based animations in combination with Spine, meaning 10+ poses per character are getting meshed, and it really eats up a lot of time to do this manually.

Erika

Hello RuinerXL. I am well aware of how the likes of Photoshop and After Effects generate casual triangles for deformation, but considering their triangles aren't then elaborated at runtime, it doesn't matter however many are placed there, yet if we were to do the exact same in Spine, the amount of suboptimal vertices created that you'd also need to counter-balance with weights or direct deformations would make your game run in performance issues pretty quickly.

Hand-crafting good meshes means you can also get away with very few points in the crucial places, something that takes little time compared to adjusting auto-generated content.

You might want to consider the use of linked meshes for meshes that are very similar across different skins: Meshes - Spine User Guide: Linked meshes
And the fact that you can now copy-paste vertices positions from meshes to other meshes/paths/anything that has a vertex.

So far I have yet to see an AI generated mesh that is truly optimized for the specific image it has in front. I think it's better to practice good mesh creation by hand instead, and smart reuse of similar meshes.

fantasyz

Its nice to know I am not alone 😃

There is actually a little story about me requesting this feature 9 months ago.
As half of my blood is a programmer and I was handling lots of characters during that time, I sense that auto generating mesh outline for shapes is a must do for a mesh-based animating software like Spine. Making the mesh cost me lots of time. From my understanding, people choose Spine is because we want high quality animation for game with minimum efforts. Something like playing audio IMO is crossing the line and is a burden as the software has one more component to maintain against different runtime. (Not to mention creating a simple testing editor to test spine animation with audio timing in unity is just a one day work. I actually don't see the point of supporting audio from spine.) Anyway, I then posted this feature request trying to draw spine team's attention saying that this could be a selling point compared to competitor like Dragon Bone Pro (1st post). After a little discussion and feeling that this feature won't make it in the next few years, I went on to check out DBP again as it mentioned "open source" in its website. Maybe I could do it myself on DBP and use it for mesh generation then export back to Spine for rigging/animating. (In fact, I found out the editor is not open source, only certain plugin is) As I try to understand more about DBP, I found this from one of their old blog post:

That's right, they actually gave it a try way before I think of it and post my request. I don't know what algorithm they use but its not accurate enough and can be improved in my opinion. The line is too close to the color pixels. All they need to do it to push the vertex outward and do some density clean up. Here is how it is doing:

The sad part is I already get use to Spine editor. I enjoy using Spine and I don't want to change. The development of DBP seems to be slowing down as well. Maybe the group lacks funding as their software does not earn them money. Moreover its editor is not open source. I cannot change their stuff to make it work like spine. I realize that most of the features I want, some of them are available in Spine and some of them are in DBP. I as a user I can do nothing about it. All I can do is to post request hoping that one day my wish get fulfilled. (and I have lots of wishes :p)

Is this feature very far away from us? No one can tell.
Maybe one day DBP polished this feature and give spine team enough pressure ... (...I am dreaming again... :bang: )

mfedorov

I decided to make such a tool. So far, not everything is smooth, but in most cases it is very useful.

Nate

Neat tool. Looks like it doesn't keep the image within the bounds. That part can be tricky, especially with low vertex counts, which is where such a feature is most useful. I'm impressed you got so far with AutoHotkey, that can't have been easy!

FWIW, we are currently working on this feature and would like it to make it into 3.8.

mfedorov

Thank you, Nate! This is great news! I will wait for this functionality out of the box.

Nate

Here's a screenshot during development, 20 vertices placed automatically:
Loading Image
Of course vertices should not go outside the image bounds, but it's getting there!

Issue for this task:
Automatic mesh hull generation · #469 · EsotericSoftware/spine-editor

It's almost done, BTW!

belpetr

Cool. This is good news))

nicmar

Awesome! Is this feature now part of Spine? How to use it? I didn't see any button for it? 🙂

EDIT: Ahh, I just saw the TRACE button. Great, that'll save a lot of time. Thank you guys 🙂