Hi, I gleaned this from the sneak peek video [vimeo.com].
Animation
• APEX (All-Purpose Execution) architecture for fast rig creation and evaluation
• APEX-driven rigging and animation toolset within KineFX: BETA
• High-performance procedural rigging framework
• Growing set of readily available auto-rigging tools
• APEX rigging network editor with unique debugging capabilities
• Efficient packed geometry format to encode entire animation scene
• Dedicated animation environment for characters rigged with APEX
• In-viewport access to all essential animation tools
• Toolbar with animator-friendly sliders e.g. key blending and resampling
• Many UX and functionality enhancements to animation graph view
• Natively integrated motion dynamics and ragdolls
• Powerful, one-click pivot setting for fast posing
• Quick and seamless switching between IK and "fake" FK
• Flexible, animator-friendly "velcro" constraint system
• Animation bookmarks on playbar and in animation graph view
• Selection Sets with easy control and access directly in viewport
• Fully customizable python interaction and rigging controls
• Hotkey and radial menus for quick access to functionality
• Dem Bones Skinning Converter SOP
• Auto-generated MikkT tangents and normals for glTF and FBX import
• Joint scaling removal option for FBX export
• Smoothing group support for FBX SOPs and ROPs
• DCC export presets for FBX Animation Output and FBX Character Output
• ML skinning with scripts, networks and training dataset In Content Library
Feathers & Fur
• High performance, GPU-accelerated feathering framework
• End-to-end feather modelling, grooming, shading and rendering
• Procedural and art-directable with state-of-the-art viewport tools
• Fast creation of single feather template or entire plumage
• Interactive noise and clumping controls for barbs and barbules
• Finely granular controls for barb shape, distribution and orientation
• Advanced feather technology for fast packing and detangling
• Plumage setup ideal for out-of-the-box Vellum simulation
• Feather Procedural for USD rendering: Karma & 3rd parties
• Export-ready plumage for game engine deployment
• Memory-efficient feather data representation
• Shared guide grooming toolset for both feathers and fur
• Designed to work seamlessly with the new Texture Mask Paint brushes
• Fur shader with medulla support
• Muscles & Tissue: GOLD
• Franken Muscle technique for faster and more robust setup and simulation
• Fast "coupled" Vellum solve for tet-stretch with superior volume preservation
• Finer controls on muscle interconnectivity constraints
• Enhancements to muscle flexing and activation
• Easier way to attach skin to tetrahedral mesh
• Improvements to skin sliding constraints
Crowds
• Constantly expanding toolset, now more artist-friendly at SOP level
• Motion Path SOP to interactively modify one or more agent trajectories
• Greatly improved collision detection and avoidance
• Support for sticky collisions and projectile motion
• Options to control clip-time behaviour e.g. Phase and Relative to Start Time
• Ability to store clip transition graph geometry in agent definition
• Easier way to retrieve specific agent shapes with Agent Unpack SOP
• Pattern matching with Agent Edit SOP
• Agent Pose from Rig SOP compilable for faster performance
• Support for UsdSkel dual quaternion skinning
Dynamics
• Whitewater and bubbles support for existing SOP-level FLIP fluid solver
• FLIP Volume Combine SOP to mix FLIP fields of different resolutions
• Efficient multigrid preconditioner for pressure projection
• Ability to generate non-intersecting bubble structures
• OpenCL Ripple Solver SOP for shallow water and deforming 3D geometry
• Projectile and Pendulum Motion SOPs for typical throw and swing scenarios
• Fast and versatile RBD Sticky collision constraints e.g. for crowds
• Cone twist constraints with drivable Car Rig setup in Content Library
• Curved glass fracturing in RBD Material Fracture
• Much faster graph colouring for tetrahedral meshes
• Vellum wind occlusion/shadow for more accurate wind blocking
• Finer masking controls for transition effects with Vellum Rest Blend
• Ability to simulate Vellum fluids with 64-bit precision accuracy
• Interactive pin handles in Vellum Brush for easy fold, drape and stretch
• Volume sharpening of pyro fields for more realistic results
• Density-based terminal velocity for more realistic pyro such as falling dust
• Redesigned modular Karma pyro shader with convenient presets
• Seamless volume looping with colour and velocity blending in SideFX Labs
• Linear Solver SOP to perform many useful linear algebra operations
• OpenCL Snippet operators for faster FX development akin to VEX Wranglers
• OpenCL support for binding topology attributes and groups for R/W
• Ability to write in-place VDBs with OpenCL
• Gas OpenCL node Improvements
Clouds
• Artist-friendly toolset for skybox and hero cloud generation in SOPs
• Billowy and Wispy Cloud Noise to help achieve highly realistic looks
• Cloud-driven enhancements to Attribute Noise and Volume Noise Fog
• Ability to quickly generate clouds from geometric shapes
• Cloud Clip SOP with viewport handles to shape the extent of a cloud
• Sky Box SOP: a one-stop shop to produce 3D cloudscapes very fast
• Sky Field from Map to produce 2D cloudscapes using existing image
• Dedicated Karma shader designed with modularity in mind
• Optimized Karma CPU and XPU volumetric rendering
• Easily adjustable cloud density for rendering
Solaris
• USD Material Catalog framework
• Much Improved material assignment/linking UX
• Growing support for MaterialX for shader development
• Native MaterialX/Karma catalog of handy materials e.g. skin and car paint
• Volume materials with presets for explosion, fire, smoke and clouds
• Volume Ambient Occlusion SOP to bake out volume for much faster rendering
• Ocean Procedural (foam, bubbles & real displacements) for all USD renderers
• Improved Hair/Fur Procedural for all USD renderers
• RBD Destruction LOP to transform pieces at render time via invoke graphs
• Karma Physical Sky LOP with Karma Sky Dome and distant sun lights
• Scene Doctor LOP to analyze, alert and repair problems in USD scene
• Geometry Clip Sequence LOP to simplify workflows around value clips
• Many enhancements to Snapshot/Render Gallery e.g. capturing all AOVs
• UI/UX improvements to the Scene Graph tree menu & details
• Better UX In the interactive Camera LOP state
• Light/Camera viewport guides with new custom display tech under the hood
• Distributed multi-frame and multi-shot rendering via "cloning"
• Improvements to Husk procedural framework
• Optimized Merge LOP performance
• Enhanced USD and LOP support in TOPs
• USD 23.08 and MaterialX 1.38.8
Karma CPU & XPU
• Absorption and dispersion
• Filmic tone mapping
• Automatic ray biasing
• Point cloud shader with fast lookup
• Improved rotational motion blur
• Physical sky system with dedicated dome light
• Optimized deep camera maps (DCM)
• Improved handling of internal reflections
• Advanced fur shader based on medulla properties
• Hex Tile shader to minimize edge tiling artifacts
• Room Map/Parallax shader to give appearance of depth
• All-purpose MaterialX pyro shader for smoke, fire, explosions and clouds
• Improved memory performance for subdivision surfaces: OSD 3.6
• Render Stats framework with special GUI and custom reporting capabilities
• Updated 3rd party denoiser support: Intel and NVidia
• Extensive OCIO enhancements throughout Houdini
• ACES colour space support
• Full Cryptomatte support
• Karma XPU: GOLD
• All the shared Karma CPU & XPU features above, plus:
• Robust multi-GPU architecture with failsafe out-of-core capabilities
• Countless speed and memory optimizations
• Nested dielectrics
• Instance variation
• Physical lens shader
• Rounded edge shader
• Light active radius support
• Volumes as first class objects: AOVs, LPEs, light linking...
• Efficient rendering of uniform volumes and frustum volumes
• Support for additional AOVs, including Cryptomatte
• Lighting: geo lights, portals, IES profiles, light filters, better sampling
• Ability to mix MaterialX standard surface shaders
• BSDF refactoring for much faster execution
• Efficient loading and eviction of tiled textures
• Highly optimized texture management
• USD Coordsys support
• Full colourspace support
• Deep outputs
3D Viewports
• Support for background video
• Redesigned custom guides framework
• Option to always display the name of the current node or tool
• Comprehensive set of customizable interactive HUDs in the viewport
• Complete redesign of drag-and-drop parameter HUD sliders
• More powerful and flexible hotkey override schema
• Better hotkey binding support for international keyboards
• Enhanced precision in selecting geometry
• New Vulkan engine: experimental-only in this release
PDG & TOP Task Management
• Services now available to PDG Command Chains
• Convenient, workflow-enhancing TOPS for Machine Learning training
• Programmable methods for stopping and restarting specific service clients
• Ability to write custom PDG nodes In C++ in addition to python
• Rolling checkpointing format to periodically save & restore TOP graph state
• Support for registering preflight handler to be invoked before graph cooking
• High demand Processor TOPs e.g. OpenImageIO and Filter by Attribute
• Performance and quality-of-life optimizations to Schedulers
• Ability to control maximum memory usage of task processes
• Recorded memory usage for all tasks that finished cooking
• Many UI/UX enhancements to TOP Interactive environment
• Additional API and env-var support for debugging PDG networks
• Text Input TOP to complement existing Text Output TOP
• SSL support In the Message Queue server
• Colourspace controls and codec selection menu in FFmpeg Encode Video TOP
• USD Analyze TOP to fetch metadata, dependencies and external refs
• USD Render Scene TOP to render USD file on disk or a LOP network
• Many TOP enhancements to USD Importing
Houdini Engine / Plugins / HQueue Platforms
• HEngine sample application to facilitate learning
• HEngine support for dictionary attributes
• Optimizations to attribute handling in HE engine
• HEngine plugin support for UE 5.3 with World Partitions
• UE Landscape Spline import and export
• UE Skeletal Mesh and Animation import and export
• Reference counted input system with condensed types and UI redesign
• macOS Apple Silicon Support for UE 5.2+
• HEngine plugin support for Unity's Splines package via new Spline Input Type
• Many Karma and USD integration enhancements to HQueue
• Improvements to ROPs and TOPs in HQueue e.g. platform-specific HFS paths
• Performance Monitor enhancements including startup time tracking
• Full adherence to VFX Reference Platform CY2023 specifications
• New Houdini video framework built on FFmpeg: v6.O,O
• Broader OpenColorIO support including Python bindings
• Python 3.10 default and 3.9 separate download
Geometry & Terrain
• Powerful and robust multigrid surfacing of massive point cloud data
• Point cloud reduction and normal computation
• Fast quad remesher with mirroring and guiding for non-CAD geometry: BETA
• Improved Topo Transfer SOP for texture map baking and UX
• Topo Flow SOP for 4D processing of scanned poses using optical flow
• UX and functionality enhancements to curve drawing e.g. soft transform
• Property editing of previously "stroke" drawn freeform curves
• Group Inversion SOP to create complementary group of geo elements
• Thickness measuring SOP using ray length or max fitting sphere
• Poly Hinge SOP to extrude polygonal faces around a pivot line
• Ability to flatten geometry procedurally
• Viewport Drawables and HUD enhancements
• Procedural and interactive 3D Texture Mask Paint toolset
• Ability to import texture flies as 2D volumes
• Option to use 2D volumes as input to Attribute From Map
• Ability to create vector and integer volumes
• Better gap closing/preservation in VDB from Polygons
• Support for frustum-shaped heightfields
• Ability to sharpen heightfields with Heightfield Blur
• Advanced biome creation toolset, including roads and terrain, in SideFX Labs
• ML terrain generation examples in Content Library
• ...And a foundational SOP & TOP toolset for Machine Learning using ONNX
Houdini 20 Sneak Peek Full Feature List
13386 20 11- eikonoklastes
- Member
- 396 posts
- Joined: 4月 2018
- Offline
- wanglifu
- Member
- 186 posts
- Joined: 2月 2017
- Offline
- SWest
- Member
- 313 posts
- Joined: 10月 2016
- Offline
- kodra
- Member
- 373 posts
- Joined: 6月 2023
- Offline
- animatrix_
- Member
- 4729 posts
- Joined: 2月 2012
- Offline
kodra
What is APEX tho. Is it something well-known that I just happen to not know, or something new from SideFX?
It's SESI tech with a cool name.
Senior FX TD @ Industrial Light & Magic
Get to the NEXT level in Houdini & VEX with Pragmatic VEX! [www.pragmatic-vfx.com]
youtube.com/@pragmaticvfx | patreon.com/animatrix | pragmaticvfx.gumroad.com
Get to the NEXT level in Houdini & VEX with Pragmatic VEX! [www.pragmatic-vfx.com]
youtube.com/@pragmaticvfx | patreon.com/animatrix | pragmaticvfx.gumroad.com
- LukeP
- Member
- 374 posts
- Joined: 3月 2009
- Offline
Seems like another fantastic release!
I was hoping for a bit more progress with the Vulcan viewport, simple geometry sculpting, improved direct modeling, an OpenCL accelerated sparse solver (like Axiom) and a preview of “ We are developing a next generation, procedural image processing system to provide the backbone for traditional compositing, texture synthesis, realtime 3D preview, and motion tracking.” technology. Or maybe even some awesome pre-rigged facial animation technology with audio recognition.
Understand that we can’t get everything at once :-). This is a massive release though, so many goodies here. Congrats SideFX.
I was hoping for a bit more progress with the Vulcan viewport, simple geometry sculpting, improved direct modeling, an OpenCL accelerated sparse solver (like Axiom) and a preview of “ We are developing a next generation, procedural image processing system to provide the backbone for traditional compositing, texture synthesis, realtime 3D preview, and motion tracking.” technology. Or maybe even some awesome pre-rigged facial animation technology with audio recognition.
Understand that we can’t get everything at once :-). This is a massive release though, so many goodies here. Congrats SideFX.
- Andr
- Member
- 899 posts
- Joined: 2月 2016
- Offline
- elovikov
- Member
- 130 posts
- Joined: 6月 2019
- Offline
- Andr
- Member
- 899 posts
- Joined: 2月 2016
- Offline
- • OpenCL snippet operators for faster FX development akin to VEX Wranglers
• OpenCL support for binding topology attributes and groups for R/W
I really hope they come with short and practical starter tutorials for learning OpenCL.
• Linear Solver SOP to perform many useful linear algebra operations
Very interesting, but I'm not sure what practical applications to expect or how it differs from a traditional Solver SOP. Any idea?
• Cloud-driven enhancements to Attribute Noise and Volume Noise Fog
More noise types in Attribute Noise are always welcome
• New Vulkan engine: experimental-only in this release
Will this help prevent UI blockage during cooking, as I've read it could be a result of adopting Vulkan?
• Powerful and robust multigrid surfacing of massive point cloud data
This is probably the single feature I'm most excited about.
Finally we can create a mesh from points.
Group Inversion SOP to create complementary group of geo elements
And this is the pinnacle of the whole Houdini 20 release. They finally created Group Invert.
• Ability to flatten geometry procedurally
I don't understand what this does.
• Ability to create vector and integer volumes
Nice!
• Advanced biome creation toolset, including roads and terrain, in SideFX Labs
I was about to say that Labs releases looked a tiny bit weak for Houdini 20, but this could be a very big feature. Very curious to test it out.
• ...And a foundational SOP & TOP toolset for Machine Learning using ONNX
Very interesting.
Edited by Andr - 2023年10月20日 09:25:54
- Carlos_Rivadulla
- Member
- 26 posts
- Joined: 7月 2019
- Offline
According Illume presentation on compile blocks, SOPs 2.0 will be exploited further in H20. However, I don't see any information about it on the list.
I'm looking forward to receiving more information when houdini goes gold, similar to what was done in H19.5, where we had some more surprises and a couple of hints about the future, just like they hinted at the arrival of vulkan, physics animation, and quadRemesher.
I'm looking forward to receiving more information when houdini goes gold, similar to what was done in H19.5, where we had some more surprises and a couple of hints about the future, just like they hinted at the arrival of vulkan, physics animation, and quadRemesher.
Edited by Carlos_Rivadulla - 2023年10月25日 01:29:36
- LukeP
- Member
- 374 posts
- Joined: 3月 2009
- Offline
- adgabrie
- Member
- 2 posts
- Joined: 8月 2017
- Offline
- ajz3d
- Member
- 580 posts
- Joined: 8月 2014
- Offline
Fast quad remesher with mirroring and guiding for non-CAD geometry: BETAThe list of new features and improvements is indeed impressive, but I'm somewhat disappointed that this particular feature has been announced as BETA. I remember that it was already supposed to arrive in 19.5, but was rescheduled to 20.0.
- Andr
- Member
- 899 posts
- Joined: 2月 2016
- Offline
- Kwesikwaa
- Member
- 20 posts
- Joined: 2月 2018
- Offline
- elovikov
- Member
- 130 posts
- Joined: 6月 2019
- Offline
Currently USD has this disclaimer: https://openusd.org/release/intro.html#not-an-execution-or-rigging-system, [openusd.org] but it doesn't mean it doesn't have ambitions to overcome that. Pixar has OpenExec it's just internal stuff planned to opensource in future releases. I think in a way apex is an alternative to that with Houdini specific operations. Basically it comes with all the stuff that sidefx created through their history and converted to "compilable" form.
Rigging is the closest thing to programming and there are several "modernish" techniques which significantly reduce amount of manual work.
https://www.youtube.com/watch?v=KLjTU0yKS00 [www.youtube.com]
https://www.youtube.com/watch?v=6TB62YKHwRQ [www.youtube.com]
However this requires a lot of custom tools and there is definitely big demand in animation for embedding advanced logic in rigging. So you can express secondary motion for example and never animate tails again.
Today both major game engines have solutions for that. Unity has custom programmable constraints (so dialect of C#) and Unreal has Control Rig with more familiar graph system. But they are locked in engines, not transferrable and requires to rig in game engines. On the other hand imo they are way more flexible and usable than most DCCs. It's not like Maya with tons of custom python code.
Basically rigging became a complex beast, but there is still no answer how to express it, ideally portable.
From my understanding apex is the new context with VOPs and compilable SOPs operators combined. So you have advanced geo processing nodes and math operations in one graph. I guess it's the lowest possible level rigging system.
Even for the purest form of animation, FK with joints hierarchy, you'll have to create an apex graph. And probably the graph topology will be similar to your skeleton, so the simplest apex graph is the graph looking like your skeleton. Though nothing stops to add any other, very advanced logic to evaluation. Your rig can be a very complex program that evaluates joints transform using a lot of inputs and operators.
For me the most interesting question is still how and who executes this graph. For example Substance have sbsar files, compiled form of their graphs. Also it has Substance Player that you can use to execute these graphs. Currently it doesn't require license and Adobe is working on USD plugin so you can use substance files in Hydra. It's even available for embedding to your application, you can distribute this "Player" with your game for example and create images from graphs on user machine (mobile phones included). Though this use case is not that popular afaik.
I think apex runtime eventually will be an USD plugin so apex graphs can be embedded in USD directly and usable everywhere. Not sure if it's going to have realtime capabilities but who knows. However these "all purpose execution graphs" and let's call it "Houdini Player" can be also an answer to recurring question on how to use HDAs in realtime on user machine without houdini distribution.
APEX is definitely not limited to rigging only, it's "all purpose" after all, but rigging is the best practical use case right now. In USD group somebody made a joke that USD with OpenExec will be able to format your disk. It's just a joke, but USD is growing much further than just a scene definition format.
Rigging is the closest thing to programming and there are several "modernish" techniques which significantly reduce amount of manual work.
https://www.youtube.com/watch?v=KLjTU0yKS00 [www.youtube.com]
https://www.youtube.com/watch?v=6TB62YKHwRQ [www.youtube.com]
However this requires a lot of custom tools and there is definitely big demand in animation for embedding advanced logic in rigging. So you can express secondary motion for example and never animate tails again.
Today both major game engines have solutions for that. Unity has custom programmable constraints (so dialect of C#) and Unreal has Control Rig with more familiar graph system. But they are locked in engines, not transferrable and requires to rig in game engines. On the other hand imo they are way more flexible and usable than most DCCs. It's not like Maya with tons of custom python code.
Basically rigging became a complex beast, but there is still no answer how to express it, ideally portable.
From my understanding apex is the new context with VOPs and compilable SOPs operators combined. So you have advanced geo processing nodes and math operations in one graph. I guess it's the lowest possible level rigging system.
Even for the purest form of animation, FK with joints hierarchy, you'll have to create an apex graph. And probably the graph topology will be similar to your skeleton, so the simplest apex graph is the graph looking like your skeleton. Though nothing stops to add any other, very advanced logic to evaluation. Your rig can be a very complex program that evaluates joints transform using a lot of inputs and operators.
For me the most interesting question is still how and who executes this graph. For example Substance have sbsar files, compiled form of their graphs. Also it has Substance Player that you can use to execute these graphs. Currently it doesn't require license and Adobe is working on USD plugin so you can use substance files in Hydra. It's even available for embedding to your application, you can distribute this "Player" with your game for example and create images from graphs on user machine (mobile phones included). Though this use case is not that popular afaik.
I think apex runtime eventually will be an USD plugin so apex graphs can be embedded in USD directly and usable everywhere. Not sure if it's going to have realtime capabilities but who knows. However these "all purpose execution graphs" and let's call it "Houdini Player" can be also an answer to recurring question on how to use HDAs in realtime on user machine without houdini distribution.
APEX is definitely not limited to rigging only, it's "all purpose" after all, but rigging is the best practical use case right now. In USD group somebody made a joke that USD with OpenExec will be able to format your disk. It's just a joke, but USD is growing much further than just a scene definition format.
- Andr
- Member
- 899 posts
- Joined: 2月 2016
- Offline
- ikoon
- Member
- 211 posts
- Joined: 1月 2016
- Offline
Can we expect a new VEX "context", running per Frames (similar to Points) and per Channels (similar to Primitives) in SOPs? I know there is CHOP vex context, but a rewrite would be amazing. I am using "proxy" geometry for such operations (and it is reliable, useful, but cumbersome).
Edited by ikoon - 2023年10月24日 05:34:27
- kevinthebright
- Member
- 214 posts
- Joined: 11月 2010
- Offline
- jomaro
- Member
- 102 posts
- Joined: 4月 2017
- Offline
-
- Quick Links