Well I read the thread and it all sounds incredibly promising. One thing is very clear – we all need to see the video. If I had a video showcasing the new RBD and cloth simulator working together, passing attributes to eachother, etc., I'd have a much better argument when makin the Houdini sales pitch to my manager. A lot of times you hear about things that are in development, but you don't know for sure until you see them in action. In contrast, we all saw the Maya Fluids demo's and thought effects animation would change forever – then we looked under the hood and werent all that impressed.
Knowing Houdini, we'll have more control than the average artist needs, and having these tools will be an incredible advantage to those of us struggling to compete with Maya's out of the box solutions.
(Having a video of these new tools in action will provide some extra artillery to those of us eager to convince our boss to invest in the software.)
I can't wait!
Let it Burn
26153 24 4-
- the_squid
- Member
- 132 posts
- Joined: July 2005
- Offline
-
- MatrixNAN
- Member
- 405 posts
- Joined: July 2005
- Offline
Hey Squid,
Softimage XSI has some pretty good rigid body, soft body, fur, and cloth systems until Houdini releases DOPS. I don't know how there fluid systems compare to Maya's. Though I suspect that when Houdini releases DOPS no other package will be able to touch them because they all just slap something together with presets and don't have a well worked out solution. Thats the main reason I love Houdini so much. It was well thought out from the ground up and continues to be. The DOPS looked awesome better than anything I have seen. Not to mention you could tell the simulation where you wanted it to end up and it would generate a simulation that would make that event occur. If it is as good as they were boosting then I suspect it would even make that paper about the simulations art directed moving from realistic to a more unrealistic art based effects by blender the two at various percentages based on users input. I can't remember the name of the paper at the moment. Sorry. You can create and define your own solvers also. The DOPs is a node design system just like the rest of Houdini. Where are you getting the $24,000 figure for Houdini because Master only costs $17,000 and that includes everything. Are you accounting for certain plugins and what not?
As far as solutions out there for fluid dynamics well martian labs has the Hydroz tools.
http://martian-labs.com/ [martian-labs.com] Just click on Hydroz Toolz.
and Syflex has the Cloth Solvers for Houdini:
http://www.syflex.biz/buy.html [syflex.biz]
http://www.pluginz.com/news/1504 [pluginz.com]
Cheers,
Nate Nesler
Softimage XSI has some pretty good rigid body, soft body, fur, and cloth systems until Houdini releases DOPS. I don't know how there fluid systems compare to Maya's. Though I suspect that when Houdini releases DOPS no other package will be able to touch them because they all just slap something together with presets and don't have a well worked out solution. Thats the main reason I love Houdini so much. It was well thought out from the ground up and continues to be. The DOPS looked awesome better than anything I have seen. Not to mention you could tell the simulation where you wanted it to end up and it would generate a simulation that would make that event occur. If it is as good as they were boosting then I suspect it would even make that paper about the simulations art directed moving from realistic to a more unrealistic art based effects by blender the two at various percentages based on users input. I can't remember the name of the paper at the moment. Sorry. You can create and define your own solvers also. The DOPs is a node design system just like the rest of Houdini. Where are you getting the $24,000 figure for Houdini because Master only costs $17,000 and that includes everything. Are you accounting for certain plugins and what not?
As far as solutions out there for fluid dynamics well martian labs has the Hydroz tools.
http://martian-labs.com/ [martian-labs.com] Just click on Hydroz Toolz.
and Syflex has the Cloth Solvers for Houdini:
http://www.syflex.biz/buy.html [syflex.biz]
http://www.pluginz.com/news/1504 [pluginz.com]
Cheers,
Nate Nesler
-
- the_squid
- Member
- 132 posts
- Joined: July 2005
- Offline
I could be wrong but I believe the $24,000 is about right.
Houdini Master: 17,000
Service Package: 4,000
Sales tax: 1,732.50
Okay… a little off. It's more like $22,732.50 or somewhere around there. The Houdini service is well worth the price and puts the Maya guys to shame. Thanks for the links… when there's a link to some video showcasing DOPs please post it!
Houdini Master: 17,000
Service Package: 4,000
Sales tax: 1,732.50
Okay… a little off. It's more like $22,732.50 or somewhere around there. The Houdini service is well worth the price and puts the Maya guys to shame. Thanks for the links… when there's a link to some video showcasing DOPs please post it!
-
- MatrixNAN
- Member
- 405 posts
- Joined: July 2005
- Offline
-
- old_school
- Staff
- 2540 posts
- Joined: July 2005
- Online
Wow, what a side-track. I thought I was bad… From fire tips to cost of software. Hmmm.
rant on…
Have any of you priced Maya support lately? Node locked pricing vs floating licenses. If I tell you as a manager that it will take 2 weeks to 2 months to do a shot, you will remember 2 weeks and I remember 2 months. If I tell you Maya complete is $1995 that is what you remember, not the floating license version which sells for a lot more and has much higher annuals. What about Maya support and turn-around on bug fixes? It is measured in years. One bug impacting 2-3 people for the duration of the project easily makes up for the cost issue.
Maya also requires much more programming resources in the larger facilities. This is a proven fact. This is required to support the maya user base which is more comfortable managing interfaces and presets than developing tools. This is a good thing and is why Digital Assets are so important to the Houdini users in large shops. Users have to be productive immediately and not prototype everything they do from scratch.
rant off…
Getting back to the original fire question, there are three things to getting fire right:
1. How many Particles and their behaviour
2. How you render the particles
3. Optimizing the rendering so you can actually deliver the shot(s) that are directable.
I have seen many different approaches to fire and smoke each with excellent results. Each one was pretty complicated. Now to exapnd just a bit on the above three points:
1. You have two basic approaches here.
- run with as few particles as you can but put the effort in to their behavour. In other words, smart particles, lots of attributes, very directable.
- run with millions of particles and keep them dumb but put the smarts in the simulation. i.e.: few to no attributes to get the speed but you need millions in order to fill the volume and get correct lighting and behaviour.
Jason's foam solution on Day at the SESI Boot at Sig shows you how to captilalize on the first option yet get the benefits of the second: run with few particles to get the right foam behavour then use these as a template to place many dumb particles with simple noise pertubation by simply increasing the gain on the noise from 0 to whatever. From 20 to 30 thousand base smart particles, amplify to millions.
Houdini currently doesn't have CFD so you have to tweak the dynamics. I feel like I am training fleas after thrashing about for a few hours with particles. There is a nice whispy POP donated from the fantastic folks at Martian Labs that has real nice dynamics feel. Check it out.
This is the only area where DOPs will help out. Only 1/3 of the solution.
2. How you render the particles is crutial to the look.
- render as 2d sprite cards from pre-rendered or real sequences
- render as dented spheres or disks with normals on.
- render as spheres with internal noise a la Gardner > Advanced Renderman 2nd edition, Creating CGI for Motion Pictures pg. 413
- render with i3d
- render with the new point clouds in H7.
- render as tons of points
- render in OpenGL hardware as sprites > motion blur? forget it.
- to motion blur or not
- etc.
Each option has pros and cons with regards to data size, render times, whether it supports proper volume lighting or not, etc. One thing novice CG artists realize is that rendering through thousands of transparent cards is awful expensive hence the third part of rendering fire.
3. Optimizing the render has a lot to do with the choices you made in 1 and 2. I have seen the most amazing 2 to 5 million particle cfd fire only to be eleminated from the pipeline by memory constrants and render times. It wasn't renderable but boy the tests even convinced Directors and Producers to throw all kinds of money at it.
It is nice to see Jason and the team at DD come up with an elegent method to scale particle rendering for volumes to simulate foam. Nice work. Both Mantra and PRMan have the ability to render points very fast making Jason's approach even more feasible at render time.
All in all, realistic fire is not easy yet as it involves many choices that burn you in the end. It is awfully expensive to do physically correct so hacks and cheats will prevail for the near future. As well, cfd solutions are only part of the solution.
Do you think production houses will give their hacks and cheats up that easily?
rant on…
Have any of you priced Maya support lately? Node locked pricing vs floating licenses. If I tell you as a manager that it will take 2 weeks to 2 months to do a shot, you will remember 2 weeks and I remember 2 months. If I tell you Maya complete is $1995 that is what you remember, not the floating license version which sells for a lot more and has much higher annuals. What about Maya support and turn-around on bug fixes? It is measured in years. One bug impacting 2-3 people for the duration of the project easily makes up for the cost issue.
Maya also requires much more programming resources in the larger facilities. This is a proven fact. This is required to support the maya user base which is more comfortable managing interfaces and presets than developing tools. This is a good thing and is why Digital Assets are so important to the Houdini users in large shops. Users have to be productive immediately and not prototype everything they do from scratch.
rant off…
Getting back to the original fire question, there are three things to getting fire right:
1. How many Particles and their behaviour
2. How you render the particles
3. Optimizing the rendering so you can actually deliver the shot(s) that are directable.
I have seen many different approaches to fire and smoke each with excellent results. Each one was pretty complicated. Now to exapnd just a bit on the above three points:
1. You have two basic approaches here.
- run with as few particles as you can but put the effort in to their behavour. In other words, smart particles, lots of attributes, very directable.
- run with millions of particles and keep them dumb but put the smarts in the simulation. i.e.: few to no attributes to get the speed but you need millions in order to fill the volume and get correct lighting and behaviour.
Jason's foam solution on Day at the SESI Boot at Sig shows you how to captilalize on the first option yet get the benefits of the second: run with few particles to get the right foam behavour then use these as a template to place many dumb particles with simple noise pertubation by simply increasing the gain on the noise from 0 to whatever. From 20 to 30 thousand base smart particles, amplify to millions.
Houdini currently doesn't have CFD so you have to tweak the dynamics. I feel like I am training fleas after thrashing about for a few hours with particles. There is a nice whispy POP donated from the fantastic folks at Martian Labs that has real nice dynamics feel. Check it out.
This is the only area where DOPs will help out. Only 1/3 of the solution.
2. How you render the particles is crutial to the look.
- render as 2d sprite cards from pre-rendered or real sequences
- render as dented spheres or disks with normals on.
- render as spheres with internal noise a la Gardner > Advanced Renderman 2nd edition, Creating CGI for Motion Pictures pg. 413
- render with i3d
- render with the new point clouds in H7.
- render as tons of points
- render in OpenGL hardware as sprites > motion blur? forget it.
- to motion blur or not
- etc.
Each option has pros and cons with regards to data size, render times, whether it supports proper volume lighting or not, etc. One thing novice CG artists realize is that rendering through thousands of transparent cards is awful expensive hence the third part of rendering fire.
3. Optimizing the render has a lot to do with the choices you made in 1 and 2. I have seen the most amazing 2 to 5 million particle cfd fire only to be eleminated from the pipeline by memory constrants and render times. It wasn't renderable but boy the tests even convinced Directors and Producers to throw all kinds of money at it.
It is nice to see Jason and the team at DD come up with an elegent method to scale particle rendering for volumes to simulate foam. Nice work. Both Mantra and PRMan have the ability to render points very fast making Jason's approach even more feasible at render time.
All in all, realistic fire is not easy yet as it involves many choices that burn you in the end. It is awfully expensive to do physically correct so hacks and cheats will prevail for the near future. As well, cfd solutions are only part of the solution.
Do you think production houses will give their hacks and cheats up that easily?
There's at least one school like the old school!
-
- Quick Links