Can we split this thread regarding animation? I'm really interested in doing characters in Houdini as we've yet to touch that aside from a few CHOPs-driven doves. But upon evaluation, we believe that it is not that abysmal platform that everyone makes it out to be.
Thanks Jordi for all the wonderful insight. -Lu On Mon, Mar 31, 2014 at 6:21 AM, Jordi Bares <jordiba...@gmail.com> wrote: > On 31 Mar 2014, at 13:44, Raffaele Fragapane <raffsxsil...@googlemail.com> > wrote: > > Ultimately I can do the same things in all three packages, in Maya in > example I don't even try to find workarounds, whenever I bump into one of > the innumerable gaps I just write my way out of it with a node, which > incidentally is also why I'm taking a looking to splice and looking forward > to their CUDA implementation instead of using my own in c++. > Text is tremendously expressive, if expensive in terms of learning curve, > which is also a cookie point for vex really. > > The problem comes when you have deadlines and you simply want to > experiment without redoing at the end of the process. For that Soft was > simply the perfect storm. > ICE limitation of having a strict I/O domain and the sequential stack with > entry points, the clarity and abundance of atomic nodes, and a generally > cohesive experience remain unbeaten.In Soft when you hit a wall you often > hit it hard, but those are few and far between, and in between you could > really fly. Same goes for clusters, properties, drag'n'drop and how Soft > presents and links those larger aggregates, they simply work 99% of the > time. > > Very true, they really hit the right spot and there is no match yet... > > > Maya and Houdini simply don't provide that experience, and their learning > curve to reach that level of fluidity is measured in years, while with Soft > we had people who never used it literally flying around within a month. > > In my experience is quite a different problem... > > - with Maya you hit a wall and that is it, either you program C++ and are > good at it or forget it... > - with Softimage you sometimes (Rerely) hit a wall, but when you do, again > there is no way out other than programming it yourself in C++ > - with Houdini is like going on the internet, is so vast you get > distracted and unless you are very focused you can be enjoying yourself > without getting anywhere but you rarely will have to program C++ unless you > are refining something for pure performance. > > As a creature TD Houdini simply doesn't get you on the zone quickly > enough, if ever. It's brilliant for a number of things, infinitely > powerful, has best of breed solvers, but it gets in the way constantly. > > My feeling is that it is too granular for many tasks and you have to be > disciplined or you can be wondering around... > > It's patently obvious they rarely, almost never in fact, had to address > teams like the ones I run as user base. > > Could you develop further? > > Performance in general is also pretty abysmal (was, might be better know) > and optimisation is opaque and lacking in immediately useful tools and > diagnostics. > > in version 12.5 and then in 13 there were some major improvements as they > embarked in a huge task to make the nodes fully multi-threaded (still in > progress) and there have been a major effort to integrate python really > well (to me feels like the best integration so far) > > Also they started to integrate OpenCL > > http://www.sidefx.com/docs/houdini13.0/news/13/opencl > > > a good example is Pyro > > http://www.sidefx.com/index.php?option=com_content&task=view&id=2123 > > Again, as of two and half years ago. Might be different now and I wouldn't > know, but nothing I've read or seen suggests so. > > Have a go a this fast rig and let me know what do you think > > > http://www.sidefx.com/index.php?option=com_forum&Itemid=172&page=viewtopic&t=31169&sid=9a764f3fbadfb00725638e42897932cf > > I have been doing a fair amount of rigging lately and I managed to put 170 > characters on a heavily choreographed scene and it was much better than > before so although is not my dream scenario it is perfectly usable and I > can do quite a few really amazing things with the rig and assets. > > hope it helps. > > jb > > > On 31 Mar 2014 23:18, "Jordi Bares" <jordiba...@gmail.com> wrote: > >> Slow performance depends on many things like having nested assets, and >> yes, you won't find an interface to manage your blend shapes but what you >> can do with your rig imho is truly phenomenal. >> >> Regarding the deformations ICE versus VOPs I would love to know more >> about it, what do you feel you can do in ICE you can't in Houdini? >> >> Assuming you are doing with the off-the-self toolkit and without any >> proprietary pipeline tools to speed up rigging building a proper asset >> interface, protect it from the user and all that takes time but do you feel >> is much longer than any other package? >> >> jb >> >> On 31 Mar 2014, at 12:04, Raffaele Fragapane <raffsxsil...@googlemail.com> >> wrote: >> >> I have to admit to not having tried again in at least two and half years, >> but I haven't seen any related release notes related to rigging since then, >> so bear with me if this is not recent or still actual information, but in >> what way is rigging in Houdini phenomenal? >> It's a major pain in the arse, generally slow both performance wise and >> to actually produce the rigs, and it has absolutely zero adequate >> facilities for a lot of stuff such as shape manipulation, and while VOPs >> are great, when it comes to deformations they don't even scratch the >> surface of what ICE can do. >> And while it's true assets are phenomenal, the sheer scope of investment >> to wrap a character up to give it to an animator is staggering, it takes >> forever to truly and properly armor up a rig and expose only the right >> context in the right way. >> >> >> >
