FumeFX v1 out

[panthon]

We made the ad by rotoscoping over some dv footage of people playing rugby and made the 3d character emitters with fume.

Excuse me for being too curious but im interested in knowing about the process. You mean u put the Dv footage as a background and create some 3d models with a skeleton which ressemble the players and match the moves?
Any chance to see some of the workflow in detail or is it supersecret?

Thanks!

[AJ Jefferies]

Fantastic job John.

[Neilg]

two things. expect to eat hard disk space at a frightening rate,

Just out of curiosity, what kind of result/scale does a 400mb sim file give you? Is that from setting an entire city on fire, or just lighting a match?

Is there a demo on its way yet? i still want to play!

[jacksc02]

it depends on a lot of things......

[joconnell]

Excuse me for being too curious but im interested in knowing about the process. You mean u put the Dv footage as a background and create some 3d models with a skeleton which ressemble the players and match the moves?
Any chance to see some of the workflow in detail or is it supersecret?

Thanks!

In terms of how it was done, we shot players on a dv camera doing the actions for the ad which an editor assembled into a cut. We took those shots and made 3d rugby players from a generic man model. We had all of the boots, jerseys and so on with the intention of doing cloth sims on the clothes for extra detail but it wasn't needed in the end - the fume sim wouldnt be sharp enough to use that kind of detail. We rigged the model up and rotoscoped over the dv footage as a viewport background in XSI and max (I'm the only max op in a softimage house but I needed backup on the roto to get it done in time). Where we used XSI we used a plugin called point oven which bakes animation to vertex animation so it allowed us to take the skinned animation from XSI into max with no rotation issues from the XSI rig - very like point cache. I did roto on some of the shots with biped too.

From there we moved into max and fume and set the animated player meshes to be fume object sources so they emited smoke. This gave us our main billow pass which rendered in brown and cream colours. I also did a particle pass for the smaller, granier bits of detail which was pflow. I used the players meshes as position objects and birthed a load of particles on their surfaces with "lock on emitter" and "inherit objects movement" turned on so that the particles stuck to the meshes as they moved. These particles weren't renderable. I then did a spawn by travel event so that my stuck particles left trails behind them and these were set to render as little mobtion blurred white cubes. These guys were sent through the fume sim with the fume follow operator and used similar movement to our billowy cloud pass.

Those two were comped over the backgrouns with some of the dv footage mixed back in for definition and that was it!

[panthon]

In terms of how it was done, we shot players on a dv camera doing the actions for the ad which an editor assembled into a cut. We took those shots and made 3d rugby players from a generic man model. We had all of the boots, jerseys and so on with the intention of doing cloth sims on the clothes for extra detail but it wasn't needed in the end - the fume sim wouldnt be sharp enough to use that kind of detail. We rigged the model up and rotoscoped over the dv footage as a viewport background in XSI and max (I'm the only max op in a softimage house but I needed backup on the roto to get it done in time). Where we used XSI we used a plugin called point oven which bakes animation to vertex animation so it allowed us to take the skinned animation from XSI into max with no rotation issues from the XSI rig - very like point cache. I did roto on some of the shots with biped too.

From there we moved into max and fume and set the animated player meshes to be fume object sources so they emited smoke. This gave us our main billow pass which rendered in brown and cream colours. I also did a particle pass for the smaller, granier bits of detail which was pflow. I used the players meshes as position objects and birthed a load of particles on their surfaces with "lock on emitter" and "inherit objects movement" turned on so that the particles stuck to the meshes as they moved. These particles weren't renderable. I then did a spawn by travel event so that my stuck particles left trails behind them and these were set to render as little mobtion blurred white cubes. These guys were sent through the fume sim with the fume follow operator and used similar movement to our billowy cloud pass.

Those two were comped over the backgrouns with some of the dv footage mixed back in for definition and that was it!

Wow, thanks for the detailed info! You guys obviously know your stuff!
When u were talking about cloth sims to guide fume sims, i was thinking that's definitely too hardcore!
I see u were using point oven...did u try the FBX workflow?
Combining the realism u get from fume and the flexibility and speed of Pflow gives great results, as we all can see in this ad. Max guy in XSI environment, that must be interesting!

[tom182]

yeah great ad, had the pleaseure of watching it (over and over) on the giant screens at Croke Park yesterday, now I got that tune in my head all day.
Hats off to the French , roll on the English

Tom

[joconnell]

Wow, thanks for the detailed info! You guys obviously know your stuff!
When u were talking about cloth sims to guide fume sims, i was thinking that's definitely too hardcore!
I see u were using point oven...did u try the FBX workflow?
Combining the realism u get from fume and the flexibility and speed of Pflow gives great results, as we all can see in this ad. Max guy in XSI environment, that must be interesting!

Yeah the cloth sim would have been totally lost in there. FBX and crosswalk were never great for us - crosswalk will only work with the simplest of rigs where you have very little parenting or constraints - we'd import models with their arms turned the wrong way or feet not working and so on - fbx has been problematic too. The nice thing about point oven is that it takes rotations out of it so it's much more stable. The bad thing about point oven is that it doesnt bring in cameras so they have to be lined up manually.

Max guy in an XSI environment sucks - you fight for render time since mental ray is such a slow problematic renderer, and if you're out sick from work, no one can pick up your job. So you don't get sick

[joconnell]

yeah great ad, had the pleaseure of watching it (over and over) on the giant screens at Croke Park yesterday, now I got that tune in my head all day.
Hats off to the French , roll on the English

Tom

Yeah we gotta sort our restarts out...

[panthon]

Fume FX 1.0 A

Just because i didnt want to start a new thread, I post here.


"FumeFX 1.0a introduces Simulation license (SL).

With Simulation license user can use backburner to send simulation tasks to any computer that has installed Full license, or Simulation license. Before submitting a job via backburner, user has to enable SL mode button in main FumeFX UI and set output path to a valid network path"

Anyone has tried it yet?
Now, if it could be distributed...

[joconnell]

You've got two choices on it - If you have a sim license, you can open the fume ui on another copy of max. The parameters are locked but you can still press the simulate button to calculate each frame. Alternatively you can set of a render in sim mode to backburner and if the sim files don't exist, that machine will calculate them. Unfortunately since fluid is dependant on the frame that has come before it it only really works well on one computer so you couldn't really have a load of machines each doing frames.

[jacksc02]

just installed the update yesterday so today I will give a sim via backburner a go

[joconnell]

Yeah it's handy if you only have one interactive max and although it can only be simmed on a single machine, that's probably not a bad thing for differences in cpus and so on.

[panthon]

Yeah it's handy if you only have one interactive max and although it can only be simmed on a single machine, that's probably not a bad thing for differences in cpus and so on.

It can be great to test different settings at once by sending variations to every node. Too bad sim licenses aren't free.