Jump to content
LaidlawFX

Houdini 20 Wishlist

Recommended Posts

Statistically improbable until you factor in a pandemic! Houdini 19 is coming out this Halloween season!

kermit-frog.gif

TM data analysis says it should have been here this summer based on years of old...

I have to say I awoke from a long slumber for this release. So I am looking forward to seeing what good offerings there are in this issue.

My only wish is for SideFX to still keep on being independent and carving its own path. Trying to get folks to understand the heart of Houdini is no small feat.

As always, this is a non-official wish list. So please submit BUGs, RFEs, and Questions directly to support. Also for anecdotal questions and rants branch them to another thread. People will more likely help you in a dedicated thread.

-Cera ;)

  • Like 2
  • Haha 1

Share this post


Link to post
Share on other sites

Animation and rigging toolset to rival Maya so I can dump Autodesk for good!

  • Like 3

Share this post


Link to post
Share on other sites
On 9/21/2021 at 11:00 PM, LaidlawFX said:

..

My only wish is for SideFX to still keep on being independent and carving its own path. Trying to get folks to understand the heart of Houdini is no small feat.

..

my only wish every-day- .... and -version

 

.cheers

  • Like 2

Share this post


Link to post
Share on other sites

From the grooming toolset there are many things i would like to see changed or added. 

Hair Generate SOP - This sop in my opinion is not very efficient because one, its a HDA. And most of the grooms i work on easily can reach in the millions. A pure rewrite of this node to be able to deal millions of primitives would be ideal. 

Two, it is doing two jobs instead of one. Its main focus should be to generate primitives and give them controls to interpolate them. Which it does have... but it also deals with distribution. This makes the process cumbersome when layering maps. Say you have a general density map for the face, you like the way everything is.... but you realize you need more density in the nose. How would you add it without changing the general seed of everything. You can make another hair generate just for the nose and merge them... this is inefficient. 

I made a HDA where you essentially can merge scatter nodes together (one for the face, one for the nose), that will be your density, run them through a single interpolate and you have your hair. Only once will you need to generate primitives. 

hairClump SOP - This node crashes a lot. I have a tutorial on my site that shows users how to control the profile of the clump in different areas without needing multiple clump nodes with different profiles and blending them together. This process when scaled up causes crashes. It utilizes Guide attributes on the tightness parameter.  Also this is another node i believe needs to re-written from the ground up. When working with grooms, it is also extremely slow to process and i have a pretty decent machine. 

guideProcess SOP -  Frizz Mode - in other packages known as scraggle or noise, is pretty much unusable. It produces very repetitive patterns and takes a lot of work to breakup. Extremely low values are need to get anything that looks like natural and even then the effect has to be toned down a lot. 

guideProcess SOP -  Displacement Mode - When dealing with grooms in the millions, the guide process node is another HDA, which means its gonna be slower than something that has been optimized. If you are displacing large grooms it would be ideal to do so in an effcient way because that kind of processing can be expensive. 

guideMask SOP - This node is very usefull, but again suffers from the problem of efficiency. It is always going to be dealing with millions of primitives and so optimizing is something that would be greatly requested by myself and other artists.  

attributePaint SOP - Options for inverting, or copying maps would be very useful.

ViewPort - I can get specular in the viewport only sometimes at work. (when i open a file, but once i start editing then it breaks). All the time when im at home. Not sure what the cause of this is, but even with the specular still isnt as good as Xgen display. 

shotSculpting hair - I have seen various solutions in the process of shot sculpting hair, i would love to have a proper system on achieving this.  

___________________________

I will be submitting these to the official RFE directly to support

 

  • Like 2

Share this post


Link to post
Share on other sites

they already started to rewrite some of that hda hair stuff into C++ in 18.5. Hairgen::2.0  and HairClump::2.0 already have their haeviest parts as c++ nodes. They become something like 10x faster after that. But Clump 2.0 is also buggier than legacy one. Lets hope they've been cleaning that up for 19, and maybe rewrote some other nodes too.

Share this post


Link to post
Share on other sites

Things that would make life easier:

1. Reorder multiparam blocks (like this)

2. Bezier curves on ramps

3. VEX multi-dimensional arrays (arrays of arrays...)

4. Solaris network snapshot gallery in SOPS

5. Upgrade COPS so that Houdini can compete with Substance Designer...

6. Destructive modelling suite

7. Fix Touch Designer integration

  • Like 2

Share this post


Link to post
Share on other sites

- Match Topology node, add the Match Vertex Order option

- Move curve seams to selected point id

- In the sort SOP better sorting option acording to bbox, with sort in X,Y,Z or from a referenced starting position but with curves continuity preservation and not just according to distance.

- PolyLoft with quad topology

- add an option in the skin SOP to realign curves profile automatically when lofting cv with same point number

Edited by sebkaine

Share this post


Link to post
Share on other sites

Better curves... Just this one thing. Please! Selecting point by point is such a pain :D Having multiple curves in one node, easily selected by box select is dream come true.

edit: oh and more blender integration would be nice!

Edited by mikuspikus
forget blender ;)
  • Like 1

Share this post


Link to post
Share on other sites
23 hours ago, mikuspikus said:

Better curves... Just this one thing. Please! Selecting point by point is such a pain :D Having multiple curves in one node, easily selected by box select is dream come true.

edit: oh and more blender integration would be nice!

 

What kind of Blender integration do you have in mind?

Share this post


Link to post
Share on other sites

Just HDA in Blender. There are some old community made solutions like https://github.com/eliemichel/HoudiniEngineForBlender and there's even a Blenders renderer inside Houdini (how cool is that?! :D)

I understand that there are problems with licensing in Blender, but it can be done and i can wish :)

 

Oh and one more tiny thing - setting pivots in fbx rop output! The Path Attribute added earlier is an awesome change. In games, pivots are very important when it comes to rendering and sorting things for the camera.

Share this post


Link to post
Share on other sites

Overlay network editor would be nice to have built-in in H19 or 20 :)

I have been using it for more than 2 years in production, so the video I initially uploaded is severely outdated. Will update once H19 is released.

 

  • Like 5

Share this post


Link to post
Share on other sites
On 8.10.2021 at 12:37 AM, mikuspikus said:

Just HDA in Blender. There are some old community made solutions like https://github.com/eliemichel/HoudiniEngineForBlender and there's even a Blenders renderer inside Houdini (how cool is that?! :D)

There is Houdini engine with python api. That should solve the license issue's. Integration shouldn't be that hard.

https://www.sidefx.com/docs/houdini/hapi/

 

  • Like 1

Share this post


Link to post
Share on other sites

- pyro sparse gpu support

- karma XPU out of beta (or out of alpha)

- some type of material point method like disney uses it for their snow stuff in frozen

- viewport improvements

- some type of new realtime engine for the viewport kinda like blender has it (it's called eevee I believe)

Share this post


Link to post
Share on other sites
On 10/6/2021 at 9:48 PM, mikuspikus said:

Better curves... Just this one thing. Please! Selecting point by point is such a pain :D Having multiple curves in one node, easily selected by box select is dream come true.

edit: oh and more blender integration would be nice!

aaaaand there it is. New shiny curves in ver 19. Thanks a lot! It's funny how happy that change made me. What have I became :D

 

On 10/12/2021 at 7:28 PM, Mandrake0 said:

There is Houdini engine with python api. That should solve the license issue's. Integration shouldn't be that hard.

https://www.sidefx.com/docs/houdini/hapi/

 

It's hard for me ;) I'm more of a hobbist-artist than a programmer but thanks.

  • Like 1

Share this post


Link to post
Share on other sites
On 10/15/2021 at 1:54 AM, mikuspikus said:

aaaaand there it is. New shiny curves in ver 19. Thanks a lot!

Probably best to reserve praises for when we'll have had the chance to do some testing.

And the same for criticizing.

Edited by McNistor
  • Like 1

Share this post


Link to post
Share on other sites

Ripple Solver 2.0 with

- support for Ripple source object with SDF collision and ability to better inject @v in the grid, 

- better physical behavior for ripple self collision 

- enhance and better connect the Ripple SOP deformer to the solver be able to art direct more precisly the ripple behavior

- ability to control the behavior at the boundary ( bounce / no bounce / continue on opposite boundary )

- integrate the tool smoothly with heightfield workflow

 

Edited by sebkaine
  • Like 3

Share this post


Link to post
Share on other sites

- I really wish that Houdini Matches Substance Designer like completely in terms of texture. I think that should be easiest since its procedural system, and also that underlying structure is already there in COPS so like COPS 2.0. but with

fundamental nodes (seamless noises, warps....) in place. I know i will probably get a lot of criticism here like (make your own node in vopcop...etc etc), but most people in production at least in fast productions dont have time to rnd how to create these fundamental

nodes. I know that there is a guy that already did this and they are called Copstance but it would be Nice if SESI made like complete overwrite like with terrains and Hair, and Kinefx...

- Also since crowds all time favorite in terms of sculpting (Zbrush) got bought by Maxon, and who knows what will happen to it (many software died after acquisition, remember XSI), i think We should step up in terms of sculpting and create the whole system to support really dense geometries like Zbrush did. That would really tie nicely with updating the whole brush system.

I know that im asking a lot :) and this is probably not something for like next release, but rather next 3 releases :), but i think it would be a great way to draw more people into H, and not just Technical people, but i also mean artists. Because they are a huge part of community. Because Currently a lot of artist are scared of Houdini, and don't see benefit of switching because they can do general artist (non procedural stuff) in other apps 100x faster. In areas like modeling, texturing, also rigging, animation, also cloth.

- which brings me to cloth. Marvelous Designer is a great example of a software for work on cloth. such an easy interface, fast, fast results, interactivity in viewport...and that's why you wont find artist going into Vellum and trying to stitch and shape cloth there, but

rather come from MD with finished cloth and prepare it for further shot based sim work.

Dont get me wrong H is my favorite software but with acquisitions that have been going on lately (Substance, Weta, Zbrush....), i am just afraid....Honestly. ... I don't want to us to get bought. So we need to push the tempo and match those apps in their strengths

so that the whole community turns to Houdini and we don't get bought. 

- Also the viewport and more Python interactions states. The latest additions on that end i think really made nice impressions in the community overall. H viewport is really behind the competition. I mean you can setup things and all that, but one shouldn't really customizing viewport out of the box.

- More and More and More please More communications with Epic and dealing with Houdini to Unreal. Houdini Engine is there and we know how to optimize things in Houdini and then when HDA's get sent to Unreal they cook really fast in H but then take another 4,5 seconds to cook inside Unreal. I have a feeling that if we talk more with Epic staff we can make it faster. 

- UE5 and Nanite are already there, and more and more Film work is switching to Unreal...and looking at Nanite tech, its really what Houdini is all about. Clusters of high res data optimized to run in Realtime. 

In general industry is going a  lot more towards realtime and i dont mean just games but everything. Also pandemic pushed film a lot towards Series...(LOKI, Mandalorian...). So this all proves the point that tech should progress much more towards realtime. We cant miss that train.

And also another thing to mention the competition is also looking at our yard and copying stuff Blender - geometry nodes, Substance - Model graph, i think another guy also made similar thing in maya for procedural modeling...i know that all that is funny when you think of what can you achieve in H with procedural workflows, but they will progress.

 

I hope i havent sounded too pesimistic, but im really having a hard time trying to convince people in my studio to switch :) 

 

Cheers

 

 

 

Share this post


Link to post
Share on other sites

Bookmarks for the timeline! Saw something like that in Maya and think it could be a cool quality of life thing

  • Like 1

Share this post


Link to post
Share on other sites

It is still extremely challenging to do small scale elegants sims involving thin sheets of liquids with the FLIP solver and the default tool.  Actually we have the option to use surface tension that force us to use a very high susbteps between 10-16, to get clean result. The main advantage of FLIP over SPH, is that it can solve about everything with just 1-4 substeps max, so this kills the advantage of FLIP over SPH in certain scenario.  The best small scale liquid houdini artist in town like Pazuzu or melt vfx guys, both ended up to develop there own particles solver to preserve surface tension + thin sheets.

 

That would be great to have by default inside Houdini :

1 - a particle based tension solver that would be based on a solver directly plug in the particle input of the FLIP solver, mimicking the concepts describe here by using mean curvature flow to compute surface tension force allowing us to keep low substeps :

https://www.cc.gatech.edu/people/home/turk/my_papers/surface_tension.pdf

 

2 - a preserving sheets particle solver, that will be able to reintroduce particles cleverly inside the sims, to avoid gap in thin sheets area in the same idea describe in this paper would also help a lot :

https://ryichando.graphics/sheetflip/download/tvcg.pdf
 


3 - an anisotropic methods for meshing small scale liquid would be very useful , the fact that VDB constrain us to use isotropic meshing is also challenging when you want to get a very clean thin flat surface that preserve detail / curvature / continuity, like describe here  :

https://faculty.cc.gatech.edu/~turk/my_papers/particle_surfaces_tog.pdf

 

4 - i would love to see better interaction between a ripple solver 2.0 and the generated FLIP mesh / sdf, to be able to post process a FLIP sims to add additional detail ripples on top of the FLIP sims base on velocity / curvature / tension. like what is describe here :

http://www.cim.mcgill.ca/~derek/files/surfaceWaves.pdf

EDIT : Artem Smirnov has just publish an implementation of this concept. To have a tool able to do exactly what happen at 00:32s in this video inside Houdini would be really useful.

 

5 - a specific 2D FLIP solver that is specially devise and optimize to work as a thin slice of data that can work with SDF and heightfield colliders. This could be combine with the new ripple solver to post process extra detail. This would allow us to generate very detail 2D map of length(velocity), vorticity, wet influence very fast, and transfert them directly to a heightfield/sdf. ( same logic of the flowmaps tools ).

 

Generally i think a better interaction between  FLIP / Heightfield / Ripple Solver 2.0 for small scale shots is really possible and would help a lot. There are very very few area where houdini is not already the perfect weapon, but for small scale elegants sims, i have the feeling that we can do better than the actual toolset, or maybe i'm wrong and small scale elegants sims will always remain the hardest nuts to crack in FX creation. :)

 

PS: i know there are some randoms piece of codes on this noble forum to solve those problems, but a clean upgraded small scale workflow done by SESI wizzard would be ace.

 

 

Edited by sebkaine
  • Like 2

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×