An approach “Oh ! Shiny new ! Let’s replace everything with it !” is counter-productive. As mentioned, PxCloth is specialized solution, oriented on character clothing (in terms of stability, performance and artistic control over the simulation). It also works on multiple-platfroms and is already used in many games. Originally, PxCloth was developed to compete with solutions like CloakWorks Shroud and Havok Cloth, where “fat” 2.8 cloth with features like tearing, inflatables, metal-cloth, full scene collisions, soft-bodies, etc has failed (as I see it, can correct me on that). No need to adapt FLEX for character clothing, better focus on use cases, where FLEX will be more valuable - advanced environmental cloth, softbodies, deformables.
(BTW, PxCloth can handle multi-layered cloth, just check Lords of the Fallen)
So far, I see FLEX more fitting in a concept of “GPU Physics 2.0”. GPU execution only, non-gameplay affective, resource intensive. Even if DXCompute version will be finished, I don’t think consoles will be able to handle large-scale simulation, and thus it won’t be used widely.
However, maybe it would be worthwhile to create a special scaled-down FLEX version, oriented on gameplay physics (maybe with CPU execution) ?. With support for trigger events, contact callbacks, etc. For games like puzzles, physics sandboxes and such.
The reason I said that is because of a first hand experience with the current Apex Cloth implementation in UE4 where we had a fairly complex character wear a robe (simulated) on top of a tall skirt (also simulated). We had many issues with self collision and penetration. Not only that, but even something like a curtain hanging from hinges on the ceilling was having many problems interacting with the player controller and we had to use hacks to get a semi decent result.
We ideally want the player to interact with it through different ways:
1- Walking through it and have it deform/pull away correctly
2- Shooting at it and having holes created through it
3- Throwing an enemy toward it and can cause the curtain to be ripped completely from the hinges on top
4- Throwing an enemy toward it and can cause the curtain to be torn into two pieces (As in the top half of the curtain is still hanging on the hinges)
When I saw the flex cloth demos with tearing and multi-layers I got hope that maybe is solid enough to sustain all those features. Now if the new PxCloth update spoke of will solve those issues in a better way then that’s great. But it will be great to get at least a rough estimate of when it will be available so we can plan accordingly (Our game is for PC/PS4/Xbox One)
Curious to see what the asset look like (if possible)
one looks like a work for FLEX, but the interesting question is - should curtain affect movement of enemy’s ragdoll ? If not, following approach may be possible even with current version of FLEX: a separate scene is created for FLEX objects, while movement of dynamic shapes (including ragdolls) in scene is approximated with kinematic bodies in FLEX scene. So the interaction will be only one-way.
However, in case availability of DXCompute version of FLEX solver for PS4/Xbox is the key factor.
@dvd.kg
Ditch the skylight. I tried using a skylight with dynamic lighting and it looks like ****. Skylights only seem to cast shadows when they are static.
4.7.4 is the latest official version of UE4 having the double side foliage lighting. 4.8 will be out soon (at least in preview version, the branch could be downloaded in Github) and it has DistantFieldGI which is Epic solution to dynamic GI: it works fine in open scenery but it seems it doesn’t work well in closed settings and the particle lighting is not finished yet. I hope both VXGI and the DFGI of 4.8 will work in tandem one day to get the best results ever.
{Sigh, cough}. I’m afraid Turf has been delayed quite a bit, other pressing issues are taking time away from it. I’ll do my best to accelerate the process.
