My problem with ICE based solutions is more on the caching & rendering side. You have to cache each and every strandpoint. With a million hairs and a few substeps for motionblur, cache files are gigantic. A good hair system should cache only cache the guides, the actual hair (and maybe even styles like small frizz etc) should be generated at rendertime with a procedural. As far as I know this is how Yeti and even Shave&Haircut in Maya works.
2014-02-14 6:52 GMT+01:00 Mathias N <mdawn...@gmail.com>: > *) First off, it should be able to stand on it's own without ICE. > *) It should be ICE compliant > *)I toyed with the notion of a node graph like yeti's but from what i > could see of it so far, it is very linear, nowhere near as fun as ICE, so a > cool idea, for visualisation but i'd have to see a little more matured > version to be sold. > > I personally don't see much of a problem with integrating the system with > ICE, and with regards to simulation it would be very difficult to separate > while maintaining the same level of adaptability that ICE provides. The > Strand Dynamics Framework that ships with ICE is slow by design, its > limiting factor being the setup rather than ICE itself. A proper simulation > system within ICE would likely have to consist of a single closed node that > would not allow any customization beyond its inputs, but it would still be > favorable to trying to completely separate it. > > Take the Strand Collision Framework mentioned at the beginning of this > thread for instance. It was built on-top of the stock strand dynamics > framework, and separates the various phases of simulation into actual nodes > in ICE. It's decent, but attempting to keep everything separated and open > to modification proved to be a major flaw preventing it from ever really > working properly. > > *) It Should have its own unit, like a strand, guide or null, that exists > for this purpose. > > A given. > > *) UI integration. where would it live ? i'd love to see the sparsely > inhabited paint tools panel, be sectioned into 3 tabs, similar to the right > side, MCP/KP-L/PPG. > Tab 1: the original xsi paint tools, weight painting etc... > Tab 2:... maybe some sculpting tool implementations :P ? (for another time > !). > Tab 3: The new hair system panel, including as preliminary features: > > Being able to groom hair without being limited by mesh resolution is > pretty much a must. Weight maps are nice for animated effects, but their > reliance on mesh resolution is a pain. I see them as only remaining for > painting attributes that need to be animated. > > Also, sculpting tools? for hair? > > 2) A layer system like the "scene" one in the KP/L so you can have > separate hair systems on different hide-able reference-able layers, with > the addition of a slider enabling you to increase or decrees the number of > hairs shown in the viewport per layer. (this would allow you to rapidly > test incremental density simulations, on the fly without having to dive > into an ice tree or bring up a ppg ). > > So multiple layers of hair in the same point cloud? That's an interesting > thought, and certainly doable. > > 3) I'm a big fan of the M tool in SI, i love the 3 small grey icons that > appear at the bottom of the view port, (manipulate, magnet and weld) i just > wish it was customisable with even more tools ! > In this spirit i would like to have a small library of icons on the hair > system panel, icons for guide painting, and manipulation: paint/erase > grow/cut, brush/frizz,orient ...etc. i would like it to be possible to > click and drag these icons into a small tray on the viewport like the M > tool small grey icons. so you can have your own customizable loadout. > > As I am still looking at ICE integration as the holy grail here, meaning > strand guides and various attributes would be generated by the plugin and > output to ICE, where they could be utilized by any number of nodes. I > suppose the only solution that would satisfy this requirement is to build > another interface that ... interfaces with an ICE tree to make the > necessary connections. This would allow for an "artist friendly" solution > while maintaining the adaptability of ICE. > > 4) Convert To: a series of utilities that allow you to convert elements, > primitives or meshes, into hair guides; so you can choose to draw a curve > and then click "convert to" guides or strands, upon which you get prompted > to choose an existing layer to add to, or to create a layer for this new > hair element. if you choose a mesh like a rectangle, it would give you the > options on how you would like the hairs to populate the inside of the mesh, > by length? what density? etc... > > The idea of a hairMesh is a popular one, and relatively simple by design. > Integrating it with a complete hair solution is slight more tricky. I > suppose standard guide generation on the whole mesh followed by selecting a > face to serve as the base layer for a hair mesh to be extruded outwards, > with the guides within that face being replaced by guides generated by the > mesh. > > Simulation of a hairMesh would also be tricky business, but combining an > off-the-shelf solution (say, syflex) with per-strand simulation might make > for some interesting results. > > > *takes notes* > > > On Fri, Feb 14, 2014 at 3:48 AM, Sebastien Sterling < > sebastien.sterl...@gmail.com> wrote: > >> I don't feel qualified to dictate the software needs of everything and >> everyone in the industry, i might expand a bit on what i'd like to see from >> such a system. >> >> For a hair system ? >> >> *) First off, it should be able to stand on it's own without ICE. >> >> *) it should probably work on its own engine (like yeti, i assume). >> >> *) If it has its own engine, it should be open so as to offer TD's and >> Dev's the freedom and space they need to solve problems and expand tools or >> implement new tools. (this point is more a given, TD's and Dev's should >> always be facilitated in this regard). >> >> *) It Should have its own unit, like a strand, guide or null, that exists >> for this purpose. >> >> *)It should be tablet compatible, as in stylus, pen pressure, all that >> jazz. >> >> *) It should be ICE compliant ; stand on it's own yes, but that should >> not subject isolated from ICE, it's too good a tool to be ignored. Ideally, >> a relationship where ICE is less of a crutch and more of a third arm. e.g >> The proprietary Guides/strands/units can be brought into ICE as data, so if >> you where making a wood nymph, and you wanted to make flowers bloom from >> the tips of her hair: all this managed through ICE. Listening to Paul >> Smith, in his demo of FUZZ, he says that strand collisions are very slow. >> So regarding hair, ICE would be more an aesthetics effects base tool then a >> physics engine, the actual simulation being taken care of by the hair >> systems dedicated engine. so yes as much compliance between the two systems >> as possible. >> >> *) UI integration. where would it live ? i'd love to see the sparsely >> inhabited paint tools panel, be sectioned into 3 tabs, similar to the right >> side, MCP/KP-L/PPG. >> >> Tab 1: the original xsi paint tools, weight painting etc... >> >> Tab 2:... maybe some sculpting tool implementations :P ? (for another >> time !). >> >> Tab 3: The new hair system panel, including as preliminary features: >> >> >> 1) A stylus pressure segment with curve profile. >> >> 2) A layer system like the "scene" one in the KP/L so you can have >> separate hair systems on different hide-able reference-able layers, with >> the addition of a slider enabling you to increase or decrees the number of >> hairs shown in the viewport per layer. (this would allow you to rapidly >> test incremental density simulations, on the fly without having to dive >> into an ice tree or bring up a ppg ). >> >> 3) I'm a big fan of the M tool in SI, i love the 3 small grey icons that >> appear at the bottom of the view port, (manipulate, magnet and weld) i just >> wish it was customisable with even more tools ! >> >> In this spirit i would like to have a small library of icons on the hair >> system panel, icons for guide painting, and manipulation: paint/erase >> grow/cut, brush/frizz,orient ...etc. i would like it to be possible to >> click and drag these icons into a small tray on the viewport like the M >> tool small grey icons. so you can have your own customizable loadout. >> >> 4) Convert To: a series of utilities that allow you to convert elements, >> primitives or meshes, into hair guides; so you can choose to draw a curve >> and then click "convert to" guides or strands, upon which you get prompted >> to choose an existing layer to add to, or to create a layer for this new >> hair element. if you choose a mesh like a rectangle, it would give you the >> options on how you would like the hairs to populate the inside of the mesh, >> by length? what density? etc... >> >> note: if you convert to a guide, them the elements will become guides, if >> you choose convert to strand then the elements will become hair in witch >> ever layer you put them, influenced by the closest guide on that layer. >> >> >> And that's it, these are some of the things i would like to see in a >> "heavier" hair system. >> >> >> I toyed with the notion of a node graph like yeti's but from what i could >> see of it so far, it is very linear, nowhere near as fun as ICE, so a cool >> idea, for visualisation but i'd have to see a little more matured version >> to be sold. >> >> i doubt anyone will ever implement any of this, this is just a lot of >> pipe dreaming, but that is my answer. >> >> >> >> >> >> >> On 14 February 2014 00:51, Mathias N <mdawn...@gmail.com> wrote: >> >>> Sticking with the idea of a complete solution within Softimage, what >>> would your definition of heavier be? >>> >>> >>> On Fri, Feb 14, 2014 at 12:23 AM, Sebastien Sterling < >>> sebastien.sterl...@gmail.com> wrote: >>> >>>> I saw that the first day it was up Eddie >>>> >>>> And as jawdroppingly cool as it is, i have a feeling we are going to >>>> need something a little heavier to go forward. >>>> >>>> >>>> On 14 February 2014 00:01, Ed Manning <etmth...@gmail.com> wrote: >>>> >>>>> https://vimeo.com/80382153 >>>>> >>>>> >>>>> On Thu, Feb 13, 2014 at 5:10 PM, Sebastien Sterling < >>>>> sebastien.sterl...@gmail.com> wrote: >>>>> >>>>>> A new hair system is kinda over due. :( >>>>>> >>>>>> >>>>>> On 12 February 2014 16:54, Luc Girard <l...@shedmtl.com> wrote: >>>>>> >>>>>>> Hi Tim, >>>>>>> >>>>>>> Thanks for the way you mentioned our projects :) . >>>>>>> >>>>>>> I would like to chime in and say that we, at Shed, are also looking >>>>>>> at hair solutions actively. Our solution based on Kristinka does the job >>>>>>> but it is very far from being an out of the box solution and it >>>>>>> requires a >>>>>>> lot of RnD to maintain as it not supported by AD. If a problem pops up, >>>>>>> with substeps motion blur per example, we are left to our own devices. >>>>>>> We >>>>>>> usually prevail but we end up spending nearly 50% of the time on >>>>>>> technical >>>>>>> details vs real look dev. I thought you should know :). >>>>>>> >>>>>>> >From what I've seen around, Yeti seems to be a good contender but >>>>>>> I've yet to do a full production setup with it. >>>>>>> >>>>>>> Good luck in your search ! >>>>>>> >>>>>>> Luc Girard // SHED >>>>>>> VFX artist >>>>>>> 1410, RUE STANLEY, 11E ÉTAGE MONTRÉAL (QUÉBEC) H3A 1P8 >>>>>>> T 514 849-1555 F 514 849-5025 WWW.SHEDMTL.COM >>>>>>> >>>>>>> >>>>>>> -----Original Message----- >>>>>>> From: softimage-boun...@listproc.autodesk.com [mailto: >>>>>>> softimage-boun...@listproc.autodesk.com] On Behalf Of Tim Leydecker >>>>>>> Sent: 8 février 2014 07:26 >>>>>>> To: softimage@listproc.autodesk.com >>>>>>> Subject: Re: Softimage Hair options? >>>>>>> >>>>>>> Thanks for this in-depth answer! >>>>>>> >>>>>>> Personally, I´m starting to lean towards going for the trial of >>>>>>> Yeti, one reason being that I think I remember Colin Doncaster´s name >>>>>>> from >>>>>>> another maya maling list and another because of the really nice sample >>>>>>> image of a bear posted by Yolandi Meiring in a similar thread here: >>>>>>> >>>>>>> (Thread) >>>>>>> https://groups.google.com/forum/#!topic/xsi_list/2erKqUcghpI >>>>>>> >>>>>>> (Image) >>>>>>> https://groups.google.com/group/xsi_list/attach/994086131ca9460/bear_still.jpg?part=4&view=1 >>>>>>> >>>>>>> >>>>>>> Another really nice one is a proof of concept of bringing (3dsMax) >>>>>>> hair-farm into Softimage from Lee-Perry Smith, with props to Dani Garcia >>>>>>> and Steven Caron. >>>>>>> That human hairdo and it´s renderings look incredibly awesome. >>>>>>> >>>>>>> http://ir-ltd.net/hair-farm-hair-into-softimage/ >>>>>>> >>>>>>> For a Melena/Kristinka workflow using Anto Matkovic´s tools in those >>>>>>> beautiful shed projects there´s a nice clip posted by/on Lester Banks >>>>>>> >>>>>>> >>>>>>> http://lesterbanks.com/2013/05/workflow-tips-for-creating-and-grooming-hair-in-softimage/ >>>>>>> >>>>>>> -- >>>>>>> >>>>>>> I have only limited amounts of time I can spend on this and need to >>>>>>> find something that has potential to be useable for testing Redshift´s >>>>>>> hair >>>>>>> shading approach when applicable but ideally integrates seamlessly into >>>>>>> either Maya/Max/Softimage. >>>>>>> >>>>>>> The combination of Maya+Redshift is allready working very well and >>>>>>> it seems it´ll be easier to successfully migrate from simple hair/fur >>>>>>> testing to something actually looking good (using yeti). Also, yeti has >>>>>>> a >>>>>>> variety of licensing options I might find atractive at a later date if >>>>>>> tempted to actually finish something beyond spare-time doodling. >>>>>>> >>>>>>> I´d prefer Softimage but if that stuff works better in Maya, it´ll >>>>>>> be Maya. >>>>>>> >>>>>>> I suck with Max, even the fastest and most intuitive plugin can´t >>>>>>> compensate that sad fact. >>>>>>> >>>>>>> Cheers, >>>>>>> >>>>>>> tim >>>>>>> >>>>>>> >>>>>>> >>>>>>> >>>>>>> >>>>>>> >>>>>>> >>>>>>> On 08.02.2014 12:57, Stefan Kubicek wrote: >>>>>>> > Hi Tim, >>>>>>> > >>>>>>> > I've just been dealing with hair an a hamster and used the >>>>>>> built-in hair&fur of Softimage /2014SP2). >>>>>>> > >>>>>>> > A few tips about working with built-in hair: >>>>>>> > >>>>>>> > Avoid too dense meshes. It creates a guide hair for every vertex, >>>>>>> > hence dense meshes make you fiddle with lots and lots of guide >>>>>>> hair strands manually, which can be counter-productive and -intuitive. >>>>>>> > >>>>>>> > If you want to edit hair parameters on a per vertex basis (via >>>>>>> vertex >>>>>>> > colors), you need to plan ahead where exactly you want your hair >>>>>>> to be and where you want certain features (transparency, density, kink & >>>>>>> frizz) to change and over which distance/area. This is especially >>>>>>> important >>>>>>> for areas like hand and feet, as well as nose & eye lids. >>>>>>> > So, before you move the mesh into skinning/rigging, you better >>>>>>> make sure your topology works not only for animation but also for the >>>>>>> hair >>>>>>> setup you have in mind. >>>>>>> > >>>>>>> > Don't rely on the built-in style transfer functionality. It does >>>>>>> > mostly work but has a tendency to "blur" the transferred hair >>>>>>> style, even if your source and target emitter topology are the same. You >>>>>>> need to move in again and reintroduce details in the fur that got lost. >>>>>>> > >>>>>>> > If you want to simulate hair with collision don't use a subd mesh >>>>>>> as >>>>>>> > the emitter. The docs say that having hair emitted from a subd mesh >>>>>>> > cannot collide with its own emitter, so you have to duplicate the >>>>>>> > source mesh and subdivide it for real (that is, create more actual >>>>>>> > polygons) and use that as the collision object. That would still be >>>>>>> > acceptable, if it worked, which it does not. What I found after >>>>>>> > tedious testing was that any collision testing fails when your >>>>>>> emitter is a subd mesh, independent of what you have it collide with >>>>>>> (itself or another mesh), which kinda sucks and is the biggest problem I >>>>>>> ran into for which I could not find a solution. Thankfully my fur was >>>>>>> rather short and the character had a lot of secondary motion, so it >>>>>>> looks >>>>>>> alife enough (besides some problems when bending arms, which are hardly >>>>>>> noticeable in the animation in my case). From what I can tell it looks >>>>>>> like >>>>>>> collision is always computed against a simplified collision sphere >>>>>>> representation of the collision object, no matter what you set for >>>>>>> collision accuracy and shape, deforming shape, etc. It just doesn't >>>>>>> work, >>>>>>> at least not for me. >>>>>>> > >>>>>>> > Sometimes, when saving and reloading a scene some hair strands >>>>>>> (like 1 >>>>>>> > out of 1000) would suddenly stick in some random direction. I had >>>>>>> the impression that it helps to always collapse the modeling stack >>>>>>> before >>>>>>> saving, at least it never occurred again in final stages of production >>>>>>> when >>>>>>> the fur description was final and not changed anymore. >>>>>>> > >>>>>>> > Next time I will surely look at Kristinka or Melena. AS for >>>>>>> > simulation...I believe there was a Strand Simulation framework >>>>>>> with self collision introduced on softimage.tv some weeks ago that >>>>>>> looked promising, I haven't heard of ne1 using it for hair so far, mayb >>>>>>> someone else can comment on that? Would love to hear some ideas on this >>>>>>> as >>>>>>> well. >>>>>>> > >>>>>>> > Good luck, >>>>>>> > >>>>>>> > Stefan >>>>>>> > >>>>>>> > >>>>>>> > >>>>>>> > > Hi guys, >>>>>>> >> >>>>>>> >> what would be a convenient way to create,style and control hair in >>>>>>> >> Softimage, with lengths up to 10-12 inches and ideally both a good >>>>>>> >> collision model and dedicated styling tools? >>>>>>> >> >>>>>>> >> Which Softimage version would you suggest, e.g. 2014sp2? >>>>>>> >> >>>>>>> >> I´m a novice with hair and fur but would like to set up a >>>>>>> manageable >>>>>>> >> sample/test that ideally works with Arnold and Redshift. >>>>>>> >> >>>>>>> >> Is it possible to work generic or transfer results from, say Yeti >>>>>>> >> into Softimage? >>>>>>> >> >>>>>>> >> Would you recommend actually leaning towards Maya for such a task, >>>>>>> >> either going directly to Yeti or similar? >>>>>>> >> >>>>>>> >> I know those are fuzzy questions, I guess I´m actually looking >>>>>>> for a >>>>>>> >> biased answer regarding any of the various hair plugin options for >>>>>>> >> any of the major three apps, e.g. max/maya/softimage. >>>>>>> >> >>>>>>> >> Sofar, I´ve found the following options: >>>>>>> >> >>>>>>> >> hairfarm, yeti, ornatrix, shave&haircut, maya hair, maya xgen >>>>>>> "hair" >>>>>>> >> in 2014ext and the cinema4d hair options (shadowmapped). >>>>>>> >> >>>>>>> >> Admittedly, that´s a lot of options and I find it difficult to >>>>>>> bet my >>>>>>> >> time-investment onto any of them since I simply know bling about >>>>>>> pro´s or con´s. >>>>>>> >> >>>>>>> >> Cheers, >>>>>>> >> >>>>>>> >> >>>>>>> >> tim >>>>>>> > >>>>>>> > >>>>>>> >>>>>>> >>>>>>> >>>>>> >>>>> >>>> >>> >> >
