Automating CLO Export
Hello CLO community,
-
You need Marvelous Designer (MD) not CLO3D, there is no mesh > topology < reduction toolset in CLO3D so your mesh will always be tri-mesh or random quads without any VFX suitable topology at welded seams which doesn't make much sense in passing assets onto any realtime environment ? You won't be able to subd them correctly leading to many artefact issues. In other words lousy CG garment assets.
If you are in VFX you will appreciate that you need to convert the character clothing into a lower poly model > bake the high res crease textures down suitable for animation on a rigged character and weight paint the garment to the avatar or rig portions of the garment. Which needs to be done in another application as part of your pipeline workflow If you are exporting cloth collision data as a recorded mesh vertex cache data then you cannot get realtime action as that is fixed animation collision data from the incoming animation track as recorded which means you have no way of getting realtime action happening as that is mesh cache vertex data and not bound as weight interpolated objects on the avatar.
Have you researched how CLO3D and MD are used in the workflow? Sounds to me what you need is not so much a python exporter (as this is simply a materials to model issue) but people whom know how to provide the garment assets in the correct formats and on characters that have been rigged correctly both incoming and externally and then tested. eg: Pose transition such that garments do not pinch or drape in odd manner around the shoulders, underarm, crotch and that hands or trim do not interfere in simulation. No amount of python inside CLO3D is going to assist in automating that human factor process as most of that is going to fall to external processes in your character VFX pipeline. Where you will get automation benefit is by systematically reducing the workflow down into a few good processes > eg: character binding < to > garment base mesh. eg: MD character skins to garment skins. ~aka weight mesh binding via smart automated interpolation.
You also need to check out really important features like they have in MD > topology tools in MD 11 that allow you to create a quick low poly mesh that has matching seam pairs for quad mesh and then it's all important binding of the low poly mesh to the high poly garment suitable for baking out UV atlas maps for the low poly model that goes onto rigging. They also have the benefit of autograding garments (one size fits all characters) so that you can apply any garment in a single size to any character shape. This can be crucial if you need to do many changes in production due to art direction as this makes your digital wardrobe usable on any character fit (unlike CLO3D which is geared to clothing grade sizes - it does not have this important VFX pipeline feature) . CLO3D has none of these mesh editing tools, as it is generally it's own standalone realtime simulation system for use with samples, and animations are generally fixed as recorded frame tracks.
When you indicate 'realtime' you need to make sure your garment assets are geared to that endpoint, and not use a halfway house.
For colorways on materials you also need to factor in things like thin mesh verse thick mesh. Which introduces all manner of downstream trim and edge (seam ) detailing that you need to factor into the texturing baking workflows which would typically mean making sure your models and trim are well planned to work across low poly.
It's not so much a matter of just export and hope it animates > you actually need to structure the workflow such that it is repeatable with the right tools > then build your python automation around that strategy.At this point I don;t think you have done enough deep research to have a decent workflow strategy as you should maybe be looking at MD and not CLO3D !
0 -
Thank you ottoline,
I appreciate your time in going into the depths, as you mostly do. I've gained many insights from your posts in the past.
In this case, the exported data is not to be animated and is static geometry. It may sound odd, but that's what is needed at this stage. I'm well aware of the process you so thoroughly described, but thanks again.Back to automation, the Python API looks to be the same for MD and CLO. I'll dig deeper into MD and see if it provides different mechanisms in aiding such a massive export task.
Thanks again.0 -
If it's animated outside of CLO3D you still have the issue of fit on the motion files for the base character to test before you finish with editing the garment assembly - I cannot stress how important that process is. So these poses (whether animated or not > eg: morphs in MD/CLO3D) need to be tested with the garment in MD so any tweaks can be made. So in terms of basic garment geometry you still need to assign a process flow in either software for the mesh geometry. Issues like suitable subd quads and loop selects come into play so that critical areas have polymesh at resolutions where it matters (eg: elbows and knees, turned up cuffs etc). There are many garment detail areas that will never translate well so you need to run a short animation cycle across your character poses in MD to test the problem areas and address them there otherwise exports and automation become useless as you will simply have to go backwards when some garments don't work as you may think. So with static garment geometry is only useful when you place it in context > eg: range of likely poses per character.
In MD you have the PBR view mode so you can reapply the baked textures (atlas maps) to the low poly mesh to see the VR outcome with art directors on the base characters pose motions in situ before you save to batch. That process needs to be handled in MD and the edits made there so you get to see how it will look and that when you have the final item it is exactly the same. This can be done rather easily but you need to plan some of the batching out in more robust python toolsets.
If you are just batching out a few mesh objects to a static meshes (under 100) , then simply use the MD mesh fit toolset to a base avatar manikin (skin) and automate that externally via mesh weight interpolation. MD will only do this one at a time and evenj with python you are limited on how you can batch this. So you need to use industry strength tools > eg: wrap 3 by Russian 3D scanner
They have an automation pipeline that will chew through literally thousands of wardrobe items from a few base character landmarks (less than 70 that you can apply to any character set base mesh. It also has a node graph and python scripting toolset that will bash anything you can do in MD by a huge productivity margin in terms of walking away and letting it run... when doing this en' mass > see their virtual try on workflow tools.
So although you can python script in MD and CLO3D it has process flow limitations (python) that cause bottle necks with human intervention - so my recommmendation if you want to nail the batch process via python for many garments (the most I have done in one single batch is 10,000) flawlessly is use wrap 3.
0 -
Thank you Ottoline,
all your comments are valuable, as always. I'm sure to come back to them when any type of animation is required.
For now, the main goal is to establish automation over many CLO project files. Any workflow and experience you can share is greatly appreciated.Thank you
0
评论