Houdini 14 Wishlist

[LaidlawFX]

I'll hunt for the ghosting fx - there was a recent fix that is in the latest production build. H13.0.401 at least on OsX with CopySop

 

Time to test Maya2015, though last time Houdini smashed the Maya viewport on a like-for-like geo  If anyone has something to test with please upload.

If you can find it be awesome... I have been on windows 7 and 8 and it's been prevalent on window since this summer, when I've went from linux studios to windows based studios. The pattern is a lot more obvious, but just hard to package. At my best guess it has something to do with multiple network pane navigation, and possibly diving in and out with U and I shortcuts. I'm beyond stumped, and the last two studios have had the issues with about a dozen combine Houdini artist, so it's not just my workflow that's causing it. 

[Guest tar]

If you can find it be awesome... I have been on windows 7 and 8 and it's been prevalent on window since this summer, when I've went from linux studios to windows based studios. The pattern is a lot more obvious, but just hard to package. At my best guess it has something to do with multiple network pane navigation, and possibly diving in and out with U and I shortcuts. I'm beyond stumped, and the last two studios have had the issues with about a dozen combine Houdini artist, so it's not just my workflow that's causing it. 

 

 

Cool - will do.

 

Is it also mainly Wondows for everyone else too?

[mawi]

Nope. I got that problem on Linux.

My impression is that curves get stuck more often then meshes. However Ive tried to replicate the bug, without any succes.

[eetu]

I'm wondering if all this manual "Set Initial", "Set Always" business is really needed. Can it really be any sort of bottleneck these days to just set always?

The only case where it is actually functionally useful is if you have an animated parameter but still you only want to set the sim value once - is that really an even remotely frequent use case?

(Or am I missing something here?)

[mawi]

I'm wondering if all this manual "Set Initial", "Set Always" business is really needed. Can it really be any sort of bottleneck these days to just set always?

The only case where it is actually functionally useful is if you have an animated parameter but still you only want to set the sim value once - is that really an even remotely frequent use case?

(Or am I missing something here?)

But if want to advect/manipulate the field/geometry you are loading from SOPs you probobly dont want it to reset to SOP values each frame.

 

If you for example load Geometry data with SOP Geometry DOP.

Set Initial-> Loads the geometry from SOPs on initial frame and then use that geometry in the DOPS "feedback loop", updating the data from prev timestep.

Set Always -> Loads the geometry from SOPs on each timestep and copying over the data on each timestep, scraping data from prev timestep.

[LaidlawFX]

But if want to advect/manipulate the field/geometry you are loading from SOPs you probobly dont want it to reset to SOP values each frame.

 

If you for example load Geometry data with SOP Geometry DOP.

Set Initial-> Loads the geometry from SOPs on initial frame and then use that geometry in the DOPS "feedback loop", updating the data from prev timestep.

Set Always -> Loads the geometry from SOPs on each timestep and copying over the data on each timestep, scraping data from prev timestep.

 

Can't there be a smarter programming way of doing this though? I just found out the hard way that sops is smart enough to do this when caching alembic files. If there isn't an animated channel up stream of the node it only cooks the network once, and creates the alembic file one way. Where as if you just have one animated channel it will write out the file for every frame. I'm willing to bet some money that there are other ways built into the software already and other creative solutions to remove this option.

 

Also from a lack of knowledge what other Simulation packages have this efficiency control built into it like this?

Edited May 7, 2014 by LaidlawFX

[Skybar]

I kinda like the Set Initial/Set Always stuff. It gives me more control over it I think, rather than if it would automatically do something I don't want. At one moment I may want something to "Set Always", and at another time using the same setup I might want it not to. For me it is then easier to choose for myself, rather than rely on some automatic thing I have no control over.

[mawi]

Can't there be a smarter programming way of doing this though? I just found out the hard way that sops is smart enough to do this when caching alembic files. If there isn't an animated channel up stream of the node it only cooks the network once, and creates the alembic file one way. Where as if you just have one animated channel it will write out the file for every frame. I'm willing to bet some money that there are other ways built into the software already and other creative solutions to remove this option.

 

Also from a lack of knowledge what other Simulation packages have this efficiency control built into it like this?

I cant really speak for the programming solution of it.

Maybe the time dependency on the SOP node being referenced could be used.

Time dependent -> set always.

else->set initial.

That doesn´t take object transforms into account though. You might have a field or geometry without time dependency, being transformed at obj level.

 

What do you see as the problem with being able to choose, is it more a lack of documentation thing?

Edited May 7, 2014 by mawi

[LaidlawFX]

Just to play devils advocate how much control is too much control as far as Set Initial/Set Always, or as far as dop nodes them selves?

 

There is literally several hundred nodes with dozens of these controls on most of them, but on other you don't have any of these controls promoted up in digital assets. It's definitely awesome to have nth level of control, but how much control is too much? I would like to believe most users are dine and dash, and not having to spend weeks working on one sim fine controlling it so much as that looses a company lots of money in fx. 

 

Personally I would still split the levels of details of dops, create a normal level that has only stalk of several dozen nodes and then dop/ultra dop level where these Set Initial/Set Always controls are and the hundreds of microsolvers are. With all my dops networks over the years no matter how different the fx whether pyro, flip, rbd, particle, at the base you use the same node and the rest of these controls don't come into play.

 

Pretty much the principle of 3 button app style design and pandora box of controls. Which does not need to be mutually exclusive, you just put pandora box underneath the 3 button solution.

[LaidlawFX]

Further a bunch of random points on the issue:

 

1. The main things is not remove functionality. Keep the functionality just wrap it differently, soon as you loose functionality in the sake cleaning you loose.

2. Save time and create better efficiency by less button clicks.

3. Optimize based on what a user uses most often, and put those non-primary options a further step away.

4. Count your button clicks, this is the key reason short-cuts, and esp digital assets are key, if you see yourself repeating a task then at certain threshold optimize it.

5. Time is money, and if a company can't keep a profit doing our work then it's hard to keep us around, so efficiency is best.

6. Think outside the box in ui design, if an uber shader has over 3,000 controls do i need to show all 3,000 controls to the user. For instance hidden ui design is important here, people in maya are more familiar with this, with their shaders. But the same logic for this applies to all ui designs.

7. Nothing can stay the same forever. Houdini is the best in updating and fixing things every day in the whole 3-D software industry, but does not mean it is not prone to being outpaced by the industry, from some young upstarts. But also that means stuff that we are used to from a decade may need to change. A radical idea at first that changes the norm, may not be as radical once it's fully fleshed out and put into production.

8. Change is hard, for any one, especially if the package works as you want and you get good turn arounds. We all thankfully make good money off of this software for a number of reason, but the people that make the most money are the ones that can do the best turn arounds, and that keeps you around longer. Do you want to stay status quo until your turn arounds are surpassed by something/someone more efficient? Like watching the software/artist of nuke surpass shake before it died.

9. Did I say money & time are king, yet? 

10. Thankfully this is a wishlist, and on a forum... And it's not sidefx blueprint for the future. They have a bunch of good heads on their shoulder and can sift through the shit. Hopefully even through a rant, or not enough info. 

[mawi]

Further a bunch of random points on the issue:

 

1. The main things is not remove functionality. Keep the functionality just wrap it differently, soon as you loose functionality in the sake cleaning you loose.

2. Save time and create better efficiency by less button clicks.

3. Optimize based on what a user uses most often, and put those non-primary options a further step away.

4. Count your button clicks, this is the key reason short-cuts, and esp digital assets are key, if you see yourself repeating a task then at certain threshold optimize it.

5. Time is money, and if a company can't keep a profit doing our work then it's hard to keep us around, so efficiency is best.

6. Think outside the box in ui design, if an uber shader has over 3,000 controls do i need to show all 3,000 controls to the user. For instance hidden ui design is important here, people in maya are more familiar with this, with their shaders. But the same logic for this applies to all ui designs.

7. Nothing can stay the same forever. Houdini is the best in updating and fixing things every day in the whole 3-D software industry, but does not mean it is not prone to being outpaced by the industry, from some young upstarts. But also that means stuff that we are used to from a decade may need to change. A radical idea at first that changes the norm, may not be as radical once it's fully fleshed out and put into production.

8. Change is hard, for any one, especially if the package works as you want and you get good turn arounds. We all thankfully make good money off of this software for a number of reason, but the people that make the most money are the ones that can do the best turn arounds, and that keeps you around longer. Do you want to stay status quo until your turn arounds are surpassed by something/someone more efficient? Like watching the software/artist of nuke surpass shake before it died.

9. Did I say money & time are king, yet? 

10. Thankfully this is a wishlist, and on a forum... And it's not sidefx blueprint for the future. They have a bunch of good heads on their shoulder and can sift through the shit. Hopefully even through a rant, or not enough info. 

 

Well, I do agree with you and I have nothing agains rebuilding stuff to get faster turn arounds and workflows.

The initial idea was to remove the button and have H do it automaticly. Reading again I think eetu talked about the force nodes and there I do agree with him.

 

However when it comes to how you want to import your data I think its crusial to let the user choose if he/she wants it to be SOP animated or DOP simulated. I dont really see how that would work in a automatic way. But yet again Im not at programmer or ui designer.

I think the function is nicely packed in the rbd object node. Simply a switch -Use Deforming Geometry.

Edited May 7, 2014 by mawi

[Guest tar]

Further a bunch of random points on the issue:

 

1. The main things is not remove functionality. Keep the functionality just wrap it differently, soon as you loose functionality in the sake cleaning you loose.

2. Save time and create better efficiency by less button clicks.

3. Optimize based on what a user uses most often, and put those non-primary options a further step away.

4. Count your button clicks, this is the key reason short-cuts, and esp digital assets are key, if you see yourself repeating a task then at certain threshold optimize it.

5. Time is money, and if a company can't keep a profit doing our work then it's hard to keep us around, so efficiency is best.

6. Think outside the box in ui design, if an uber shader has over 3,000 controls do i need to show all 3,000 controls to the user. For instance hidden ui design is important here, people in maya are more familiar with this, with their shaders. But the same logic for this applies to all ui designs.

7. Nothing can stay the same forever. Houdini is the best in updating and fixing things every day in the whole 3-D software industry, but does not mean it is not prone to being outpaced by the industry, from some young upstarts. But also that means stuff that we are used to from a decade may need to change. A radical idea at first that changes the norm, may not be as radical once it's fully fleshed out and put into production.

8. Change is hard, for any one, especially if the package works as you want and you get good turn arounds. We all thankfully make good money off of this software for a number of reason, but the people that make the most money are the ones that can do the best turn arounds, and that keeps you around longer. Do you want to stay status quo until your turn arounds are surpassed by something/someone more efficient? Like watching the software/artist of nuke surpass shake before it died.

9. Did I say money & time are king, yet? 

10. Thankfully this is a wishlist, and on a forum... And it's not sidefx blueprint for the future. They have a bunch of good heads on their shoulder and can sift through the shit. Hopefully even through a rant, or not enough info. 

 

 

Yeah - but does anyone want to be in bed with AD after that killed the SoftImage? The management there is weak and ineffectual.  They glorify junk and your brain will rot if you drink their koolaid.

 

Overall I like Kai's Krause's description that software is performance art

http://edge.org/response-detail/10113

[edward]

Overall I like Kai's Krause's description that software is performance art

http://edge.org/response-detail/10113

 

For some reason, it never dawned on me that behind Kai's Power Tools was actually a person named Kai. Thanks for the link!

[eetu]

I kinda like the Set Initial/Set Always stuff. It gives me more control over it I think, rather than if it would automatically do something I don't want.

 

I've definitely gotten used to it a long time ago, and I guess there are cases where it is useful.

It just resurfaced as fresh people I've introduced to DOPs seem to find them cumbersome and it seems to be a source for mistakes.

 

I mean:

- I see a parameter

- I put an expression or a curve in it

- By default it does nothing

 

For a new user that can clearly be an issue.

[magneto]

I am also with eetu I think it creates visual clutter and is not elegant. IMO it would be better if there was a node that could override, set this behaviour for any parameter, and therefore eliminate it from each parameter.

 

Maybe an intelligent system where this could be detected might work as well. If there is a channel (keyframe, expression), set always, otherwise set initial. Because SOPs have the same thing but they don't allow you to disable evaluation of a parameter if there is a channel, unless you delete the channel.

[cgstar]

delete add  bounding type (bounding obj ) 3Q 

[cgstar]

delete add  bounding type (bounding obj ) 3Q 

[Guest tar]

Nope. I got that problem on Linux.

My impression is that curves get stuck more often then meshes. However Ive tried to replicate the bug, without any succes.

 

Just curves themselves or things created curves?  Also any benign hints on what anyone is doing i.e using dops, or dropping nodes by the tab key etc will help as workflows usually are the key to reproducing the ghosting issues.

 

So far I've been actively trying to break Houdini this way and can't.... normally it only takes <20min if I'm trying to

[LaidlawFX]

Just curves themselves or things created curves?  Also any benign hints on what anyone is doing i.e using dops, or dropping nodes by the tab key etc will help as workflows usually are the key to reproducing the ghosting issues.

 

So far I've been actively trying to break Houdini this way and can't.... normally it only takes <20min if I'm trying to

My setup usually has two to four network panes and I switch between them and in out of subnets pretty fast with U and I and the pane back and froth buttons, plus the go to buttons on network connections like merge, and they all trigger it at some random time. Plus I have multiple houdini sessions open, one simming or rendering and another being prepped, for the next task to go. Usually high poly counts 100,000 points or more and per frame changes, mostly sim style work. It's kind of like the individual viewport's frame buffer chokes. You can do a screen capture after it happens, but not very useful, as triggering it is the key to be able to debug it. Closing that viewport pane gets rid of it, or going back to the node that was originally displaying that geometry will get rid of it. Almost like the viewport can remember what node it comes from, but has alzheimers and forgets it's not lookign there anymore and forget to hit the dump command. So if there is some sort of buffer in the viewport per a node(like when you template a node and display another node), so I guess however that viewport info is saved chokes if you move too fast. Something similar to when you copy and paste in Houdini too fast and it does not have time to write to the copy buffer in your temp, before it runs the paste command again, pasting nodes that were originally in the buffer. It's almost like the viewport needs a refresh button(I do not want that as the solution). So definitely like explaining a ghost, you see it but you can't repeat it ever, unless you've also seen the ghost. 

[mawi]

Just curves themselves or things created curves?  Also any benign hints on what anyone is doing i.e using dops, or dropping nodes by the tab key etc will help as workflows usually are the key to reproducing the ghosting issues.

 

So far I've been actively trying to break Houdini this way and can't.... normally it only takes <20min if I'm trying to

Just curves.

Been building tools for fur and hair lately, creating modifiers for guides generated in xgen, zbrush and Houdini.

 

Maybe we should move this to another thread.