🛈 Announcement
Greetings! The Division forums are now archived and accessible in read-only mode, please go to the new platform to discuss the game
  1. #21
    The whole debate can be summed as such:

    All we have to go on is a few gameplay videos, not even analyses but just gameplay videos. By their very nature videos are 2D, this leads to a problem where this can skew the perception of a 3D object as much of the information is lost, a pose (or an animation) can look great from one perspective and at the same time being completely garbage when seen from any other.

    Relevant video on the matter:



    Make what you want of it, I personally don't see much of an issue.
    Share this post

  2. #22
    Originally Posted by Trippul G Go to original post

    You can't just load up a video of something into a 3D program, and expect to magically have 3-dimensional data. .
    Who said anything about loading video into a 3D program? My duh! moment really is, we don't have access to the TD2 models yet.
    Share this post

  3. #23
    Trippul G's Avatar Senior Member
    Join Date
    May 2016
    Posts
    1,523
    Originally Posted by III_Hammer_III Go to original post
    Who said anything about loading video into a 3D program? My duh! moment really is, we don't have access to the TD2 models yet.
    Even if you did have access to them, how would you go about loading your main character from either game, into something like 3D Studio Max, with all animation data intact (not to mention things like normal maps), so you can do your proper side by side comparison?

    I'd love to know.
    Share this post

  4. #24
    Originally Posted by Trippul G Go to original post
    Even if you did have access to them, how would you go about loading your main character from either game, into something like 3D Studio Max, with all animation data intact (not to mention things like normal maps), so you can do your proper side by side comparison?
    Would texture maps (or even character models to an extent) be all that relevant when comparing or analysing an animation? Just the armature and animation would be good enough IMO.
    Share this post

  5. #25
    Trippul G's Avatar Senior Member
    Join Date
    May 2016
    Posts
    1,523
    Originally Posted by BOT_Coyote Go to original post
    Would texture maps (or even character models to an extent) be all that relevant when comparing or analysing an animation? Just the armature and animation would be good enough IMO.
    The model is what the animation data is applied to. Normal maps are not "texture maps" in the traditional sense, but rather they are applied to a model to increase visual detail without inflating the polygon count (important for real-time work). As such, strictly speaking, they are not REQUIRED, but they CAN significantly impact the overall look of a model, and should definitely be considered.

    Now on to getting hold of the animation data. You basically have two options: bringing in the animation rig, which consists of the skeleton and control hierarchy; or bringing in point cloud data.

    But here's the rub: transferring data, especially animation data, between software packages is tricky at the best of times, due to every single package doing it slightly differently. Coordinate systems might be different ("up" in package A might be "straight ahead" in package B), entire categories of data might get dropped because they have no direct equivalent in package B, etc. This is why transferring animation data is avoided most times, if at all possible. It's generally done in one package, and that's where it stays.

    The other option, as I mentioned, is bringing in point cloud data. This is generally only used for things like smaller (low polygon count) objects, but can be applied to entire characters, if need be. Problem here is that this data can be HUGE. In it, you're storing the X, Y, and Z position of every point on the model, for every single frame of the animation in question. Say you've got a character model with 10,000 points (which is really quite low, generally speaking)...for one frame, that's 30,000 numbers. Depending on what fps you're animating for, (24 for film, 30 for TV, etc), those numbers VERY quickly can balloon out of hand and turn into some unwieldy file sizes.

    The other problem with getting point cloud information applied is that it relies on your source model and your destination model being identical. And not just "yeah, that looks the same", but point for point, polygon for polygon, IDENTICAL. When character models (or any kind of model, really) are built, they maintain something called point order, which is a sort of history of how they were built. Models are made of polygons, and polygons are made by connecting points. When you create these points and polygons, each point that is created is assigned a unique ID. The first point that is created is assigned #1, 2 for the second point, 3 for the third, and so on.

    When point cloud animation data is applied to a model, the point order of each model (source and destination) needs to be EXACTLY the same. If not, it either won't work at all, or your character will just be a jumbled mess.

    Well, this shouldn't be a problem, you say, right? You're taking the model from the game and you're going to bring him into 3D Studio Max, so everything will be the same, right? First of all, no one has any point cloud information except the developers. It's ridiculously large data, and there would be absolutely zero reason for them to include it anywhere on a game disc. So that's out.

    But back to the character model. It's not like you'd be bringing in a 3D Studio Max-native file. Sure that's where it may have originated, but again, they're not going to give you the source file on the game disc...what's there has been exported and converted into a (often proprietary) format which strips out any extraneous data and gives the game engine only that which it absolutely needs. It's like if you create an image in Photoshop. You're not going to put the original PSD file up on your web page, with all of its layers and masks and alpha channels and what not. You're going export it down into a jpg or a png or whatever. It still looks like the original, but all of that extra data is gone.

    So even if you brought the model into Max and it came in perfectly, you're still not getting the "original" data. Most times, when people try bringing game models into 3D packages, the results are less than perfect. The data that's there may need to be unpacked in a certain way, then there may be other files that need to be applied, etc. When doing this, things like point orders can get messed up, polygons can be doubled up or inverted, and all sorts of other problems. It's all generally pretty finicky because people are trying to reverse engineer something that the developers don't generally want you messing with in the first place.

    So...bottom line, after all of that, is this: even if you do manage to get your game character model into your 3D software properly, you're going to have one hell of a time first finding the animation data, and then bringing it in and applying it accurately. Unless you're a member of the dev team, I dare say that it can't be done.
    Share this post

  6. #26
    Originally Posted by Trippul G Go to original post
    Even if you did have access to them, how would you go about loading your main character from either game, into something like 3D Studio Max, with all animation data intact (not to mention things like normal maps), so you can do your proper side by side comparison?

    I'd love to know.
    I don't have to load my character from the game, just the skeleton and animations that already reside in my hard drive. What do normal maps have to do with anything?
    Share this post

  7. #27
    Originally Posted by Trippul G Go to original post
    The model is what the animation data is applied to. Normal maps are not "texture maps" in the traditional sense, but rather they are applied to a model to increase visual detail without inflating the polygon count (important for real-time work). As such, strictly speaking, they are not REQUIRED, but they CAN significantly impact the overall look of a model, and should definitely be considered.
    Thanks for the explanation and an overall interesting read but I too dabble in 3D and I am aware how most of this stuff works, it's not 15 years of experience in the field like you but I manage.

    I only wanted to suggest not to make the whole problem overly complex at the beginning, if there are concerns about the animation, I'd focus on the animation alone first before slapping geometry and textures onto it.
    Share this post

  8. #28
    Trippul G's Avatar Senior Member
    Join Date
    May 2016
    Posts
    1,523
    Originally Posted by III_Hammer_III Go to original post
    I don't have to load my character from the game, just the skeleton and animations that already reside in my hard drive. What do normal maps have to do with anything?
    Please point me to an example of someone who has bought The Division, and loaded their character model into 3D Studio Max, and who has successfully applied game animation data to it, so that the result is a 1:1 replica of what's seen in-game. Seriously, I'd love to see it.
    Share this post

  9. #29
    Trippul G's Avatar Senior Member
    Join Date
    May 2016
    Posts
    1,523
    Originally Posted by BOT_Coyote Go to original post
    I only wanted to suggest not to make the whole problem overly complex at the beginning, if there are concerns about the animation, I'd focus on the animation alone first before slapping geometry and textures onto it.
    Absolutely, I get what you're saying, and generally you're right, you want to keep things as simple as possible. But a lot of times with things like character tests, you'll want things like normal maps applied since often the geometry can look quite different if they are/are not applied.
    Share this post

  10. #30
    Trippul G's Avatar Senior Member
    Join Date
    May 2016
    Posts
    1,523
    Basically, my whole point is, if you're saying that this character looks worse than another character, logic dictates that you look at the characters in as close to their final form as is possible and practical, to determine their differences.

    In the case of comparing animation, assuming you want to do it in a 3D environment and not just look at 2D rendered videos, you don't need things like color maps, but you absolutely at least need the character model itself, along with whatever rig has been created and used to animate the character.

    You can't just compare a rig to another rig, with no character model, as, if I'm not mistaken, seems to be what you and Hammer have suggested. Assuming you have animation data applied to the rig, you'll be able to get a sense of overall timing, and that's about it. Without a character model (actual physical geometry) to go with the rig, you're essentially flying blind and are just guessing what the character will look like when the rig is applied to it.

    Originally Posted by III_Hammer_III Go to original post
    I don't have to load my character from the game, just the skeleton and animations that already reside in my hard drive.
    What are these "skeletons and animations" that already reside on your hard drive? Where did they originate from? How are you viewing them?
    Share this post