Houdini 10 Wish List

[Jason]

Hi all,

This thread is inviting everyone to write down what they'd really like to see happen in Houdini 10.0. So it seems Houdini 9.1 is nearly out the door and so let this be a forum for all your needs and whims and ideas. What does SESI need to do? Where would you like to see Houdini go, in this crazy market? Try to be clear about which wishes you consider absolutely essential and which are nice-to-haves. If you desire a feature present in other software, extend it to be better than that feature and write it down in here. Welcome to the thread were wants and needs are the same thing, really!

After we've consolidated the resultant list a little, we'll put together a little popularity poll of the main contenders. Let's make it clear to SESI what we need.

It's time to be vocal! Write early, write often.

Jason.

[vmuriel]

More threading !!!!

Now, with dual, quad, ... cpus, a good parallel schema is basic for good performance.

Especially in DOPs.

Edited January 21, 2008 by vmuriel

[Jason]

More threading !!!!

Only one thing? Come on, man - we need like 10 things from every person. Go crazy, list them in order of importance/desperation.

[Tamis]

1. Traking in COPS (2D and 3D)

2. better python intergration <-- put the whole team on that it's brilliant.

3. even better documentation (alway's gooooooood)

4. faster volume rendering, would be nice

there's my 2 cent's.

[symek]

1. Fast cloth solver + hair toolkit - so one who considers character animation in Houdini will have no excuse and could jump to work in run.

2. Voxels Mega Toolkit - so there was a decent tools to deal with voxels based modeling.

3. Layered animation: save, collect, edit, trimm, retarget, apply and combine any piece of animation data.

4. Build-in SVN integration (pretty straightforward to do with python).

5. Collada export (possible also Fbx).

6. Let's build-in shaders follow industry standards...

7. Halo - so close to be usable: tracker, keyer, painter, stability, specially viewer's speed and responsiveness!

8. Stable plugins API? So someone out there could actually start developing Houdini's plugins...

? ...ah, what a hell:

9. Foundation of character behavioral simulation: easy in/out for skeletons, ANN library, special object "skeleton", "sense", "brain".

10. Realtime engine integration (proprietary or well acclaimed).

Edited January 22, 2008 by SYmek

[MIguel P]

Faster volumes, that's the main thing I'd want. And also faster fluids.

[digitallysane]

1. Stability

2. Consistency

3. Speed, speed, speed -- everywhere

4. The long overdue integration of CHOPs to the rest of the package, with interactive tools, easy NLA etc. Development of what started with the Channel CHOP.

5. On the same lines, easy motion re-targeting, in CHOPs, and easy layering of animation

6. General fine tuning of the character tools. They are on a very good path, but there's road ahead. Faster muscles. More capturing options in the autorig.

7. Working Cloth

8. Complete hair toolset

9. Working and fast render region, competitive with what's available in other packages.

10. Faster display. Like, 10x faster.

11. Real time shaders, in VOPs and in the viewport, and also good and useful representation of Mantra shaders in the viewport.

12. Fix the mess with the materials: SHOPs? VOPs? VOPs in SHOPs? VOPs outside? Consistency, please.

13. No holes in the Python implementation.

14. Better filtering in Mantra.

15. Faster render of subdivs.

16. Subdiv representation in the viewport.

[ofer]

Better modeling tools. Houdini should not aim to replace apps like modo or zbrush, but a few features will make Houdini a very good modeling tool:

- "Smart Groups" - When modeling in Houdini, I very often choose a group of polygons for each SOP, and if I add, remove or reorder polygons above a SOP with primitive numbers in its Group param, this SOP "breaks", and I need to manually reselect the primitives. It is not always possible to select using an expression, and often quite difficult.

I want Houdini to detect these cases and update the Group parameter of every affected SOP. It is possible in most cases. This feature will make a Houdini a much much better procedural modeler.

- Loop Slice - I can't work without it. It is the feature I use most in Modo, and the main reason I don't do most of my modeling in Houdini.

- Edge Sliding

- Better Viewport Rendering - I guess it possible to use better OpenGl shaders in the viewport to aid the viewing of different geometrical features.

- Faster, more stable cloth solver.

[Marc]

1. Some decent python docs for the libraries would be nice

2. Subdivision as a true geometry type (faster)

3. Faster/more stable cloth

4. More shaders!

5. better integrated chops

6. More modeling tools, selection types etc. It's all been said before.

[michael]

my list part A

not in order of importance

1 ) unification of deform and muscle_deform (or at least the ability to follow one with the other)

2 ) unification of all character tools between bone workflow and muscle workflow (edit regions, edit weights etc)

3 ) view subD limit surface in viewport

4 ) full collada support

5 ) direct parameter linking (obj HDA_1 param > obj HDA_2 param > obj == obj HDA_1 param > obj )

[SpencerL]

The ability to have handles for paint,edit,etc SOPs inside an OTL

[sibarrick]

1. Improvements to the edit sop, point sliding and brush tweaking on positions. Not just peak sculpting, translation of points perpendicular to the camera.

2. Faster more stable sculpting of high res meshes.

3. Shader blending without having to build "uber" shaders first.

4. Full ui scripting - scriptable python based viewport tool set.

5. Full hair sculpting toolkit.

5. Finish muscle system and document it.

6. Better intelligent viewport selection - including pre-selection highlighting.

7. User data saveable in locked assets by storing it in the hip file. ie edit sops handles exposeable in assets

8. Nestable takes - this is probably number one on my list in terms of usefulness.

9. Layer based animation

10. Collada export.

probably more but that's for starters...

actually the more observant of you will notice I used 5 twice, so this is already 11....

Edited January 21, 2008 by sibarrick

[hoknamahn]

1. Rendering from multiple cameras in one pass

2. Blending of takes

3. Get away HScript

4. Cool feature for any kind of networks... Now we see our scene as nodes but we know that nodes are just visual representation of a code (Hscript, Python, VEX whatever). That would be great to have a switch to change the way we see a scene in a network pane "Node / Code". In "Node" mode we see and work with nodes but in "Code" mode we can write and check a code directly. Any code, VEX, RSL, Python... In any network.

5. Work with vertex normals in all modeling tools with proper visualization in the viewport.

6. Nice small button allowing to switch between component / vector mode for any 2, 3, 4 float parameter.

7. Wavelet compression for 3d textures

8. Improvements in UI. Some new features are not as good...

8. Bug fixes, bug fixes, bug fixes.

[andrewlowell]

I'd agree with many of the other posts but here's my list of 10 ...

CHOPs

1. An "Edit" CHOP that allows NLE style editing like an audio or video editing program

2. A "Python" CHOP definition like the SOP definition

3. CHOP sample rates up to or beyond 384 khz

4. An "interpolation" CHOP (not like the current one) that allows interpolation between trigger values with all of the standard interpolation types like spline, constant, etc

COPs

1. An "Edit" COP that allows NLE style editing like an audio or video editing program

2. Great keyer

3. Great tracker

DOPs

1. Support for soft-body interactions like cloth / wire collisions etc

UI

1. Audio output to Jacks on linux, and ASIO on Windows, multiple CHOP nodes selectable for surround sound / various output channels

Digital Assets

1. Unlimited inputs to DA's, 4 is never enough!

1. Rendering from multiple cameras in one pass

Use the switch operator?

[michael]

Digital Assets

1. Unlimited inputs to DA's, 4 is never enough!

I think the current visual limit is 6 - and there is no actual limit (well maybe 1024 or something...)

[andrewlowell]

WOW ... is this new in 9 or have I always just been "dense"!!

Edit ... yup H8 too .. I'm dense ... (put this in an obvious place in the help people)

Ok I get one more wish ..

GUI

1. Ability to resize split views like in 3ds Max

Edited January 21, 2008 by andrewlowell

[Jason]

• Perform (modeling) operations on multiple objects simultaneously - handles to be able to bound to multiple objects. Essential for hard-surface modeling with multiple objects. We have to attract people to the Houdini modeler so that they aren't forced to buy Maya/etc and thus risk never returning.
  
• Vertex normal support throughout Houdini proper. Houdini will always have an uncomfortable fit in a pipeline alongside other 3D software until this is solved.
  
• Thread all solvers more (esp. PLS fluids) and support sending coupled fluid simulations to multiple machines to make fluids more usable in production.
  
• Volumes from VEX : create SOP to create geometry volumes from Image3D shaders, and a CVEX SOP to be able to draw voxels in Volumes.
  
• Improve viewport picking and handles (snap-along-axis[align], point slide, grow/shrink selections)
  
• Reintroduce and improve IPR and auto-archiving, perhaps introducing a resident mantra with Hal Bertram/FPrime-like resolution enhancement and..
  
• Make Mantra's PBR renderer production ready : increase speed, allow deep rasters to be generated, solve the problem of splitting out components for compositing (glossy, specular, diffuse). Together with IPR, this could be an awesome tool.
  
• Make a workflow to pick and assign shaders on a geometry level : Allow picking of primitive groups while maintaining group name. A test for this - try to effectively material up a heavy composite object - say a Transformer. Right now it's so painful I'd call it impossible, practically speaking.
  
• DOPs workflow needs to be straightened out by someone who is not a programmer nor an intern. There are a lot of gotcha's, accidental omissions and snafus that harass the process.
  
• Improve Network Editor to include all the cool productivity features we find in the likes of Nuke, Shake and Fusion. Wiring/Post-its/Alignment/Dot-nodes/etc.
  
• Permeate HOM throughout more parts of Houdini. Allow HOM callbacks in event loops. All UI gizmos should be routed through HOM. Handles should be able to be intercepted by HOM. We can eventually have a textport "command echo" like in Maya.
  
• Viewport to attempt to render texture swatches using the VEX engine. Viewport to render shadowmaps to do OpenGL lighting. The viewport is currently absolutely useless wrt any texture/lighting work. It's plainly utilitarian. This would be more useful and faithful to Mantra rendering than the introduction of VOP-based OpenGL2 shader authoring IMHO, although I can see game developers wanting this as a counter-measure to Mental Mill.
  
• ImagePaint SOP: a PaintSOP which works like a 3D paint tool - which applies brushstrokes to a 2D map along uv's instead of point attributes. Use the Gimp renders for the paint-strokes, perhaps? Image data should be able to saved both in the hipfile as blind data, or manage external image files.
  
• Viewport needs to be able to visualize fields better. In particular it needs to visualize SDF's and vector fields.
  
• All those brilliant DOPs routines which can be need to be made available to SOPs - would love to be able to run advection, surface tension, pressure calcs, move to iso, the stuff from Gas Calculate to go into Volume Mix. There is a watershed between SOPs and DOPs that is screaming to be broken down.
  
• Little ones: draw crenellations/marks on ALL sliders to indicate values - especially zero and the integers. Cache DOP frames to appear on time-line. Flipbooking drawn to off-screen buffer.
  
• Give POPs some love- in particular, adding SDF collisions to the Collision POP and direct support for using Volumes for field forces. Going via the DOPs solver architecture is slow and hard to set up and dense - which is in part due to the design of DOPs. We should be able to generate fluid sims are then use those forces easily in POP solvers - a very very common workflow in FX.
  

[Jason]

1. Rendering from multiple cameras in one pass

Well, its technically not a "pass" if its a different camera. Passes are meant to be all from the same camera. So what do you mean by this?

However! It should be technically possible to append alternate cameras to the end of an IFD and issue several ray_raytrace commands, one for every newly defined camera... and it should be able to re-render the scene and possible save on cached data (eg,irradiance cache or micropoly cache [for raytracing displacements]) or generated acceleration structures (eg. the raytracing octree). The problem is that the last time I tried this myself, Mantra crashed.

[lynbo]

In no particular order

Render Pass/Partition scheme... on steroids, things like apply materials and operations to entire "Pass/Partition"

and built in "path management", based on say project name for seq/pass/token/ect relationship.

-I know you can do this with bundles takes and different render ops, but its not as intuitive as my pea brain would like-

Definite re work of materials / texture paradigm.

really don't like in its current state.

Easy simple Sprites workflow. Drag and drop my material on particles or object and "bam" - Sprites!

lots and lots of setup presets in python very much like the current shelf buttons do, but perhaps in a different way.

I don't want any more clutter in the window.

in the tab menu for instance or in another "pane/explorer" that you can call up and dig around at your leisure.

Chops additions like animation clip mixing

the ability to move animation around as clips and save them as assets and such.

You may be able to do this now but I don't know how. I need more chops training myself.

optimized GI, final gather, caustics rendering

PBR is wonderful, but it would be nice to be able to mix and match rendering techniques as needed

This may just be my bias as a XSI mental ray user. I like that I can check GI, FG, and caustics on and off and adjust settings separately.

sim speed,

render speed,

volume speed,

interaction speed,

speed speed speed everywhere!

did I mention speed?

Blast almost forgot

implement 3D equalizer inside of Houdini.

bit much I know but some 3D tracking would be fantastic.

3D composite tools al la Nuke, Fusion, ect.

[Marc]

in particular, adding SDF collisions to the Collision POP

I second, third, fourth and fifth this one (I'm an admin, I'm allowed ). I can't believe I forgot about it, the one we wrote was instrumental on Spidey. This one shouldn't even be debatable actually...