I've seen lately some really cool tutorials about how to export CSV and Alembic data from #Houdini to #PCG graph in Unreal. I was wondering what could be done with #Python remote execution and PCG data assets ( introduced by Epic Games in 5.4 ). I've really liked those PCG data assets lately, really well integrated in #Unreal content browser, and can be read easily as parameters in PCG graphs. Turns out, with a bit of Python and #Blueprints, you can have a live link between Houdini and Unreal and send the points cloud with just one node 😁. I really think that would be super useful to have that built-in the Houdini Engine C++ plugin directly ! ( SideFX Software 😉). More info here: https://lnkd.in/ewJ5a839 #Unreal #Python #PCG #Pipelline #Dev #Houdini # HoudiniEngine #Epic
Awesome stuff! For the live syncing, have you seen Session Sync? That essentially handles the live updates while working in Houdini. The missing component would be the PCG export/bake capability
HoudiniEngine already can transfer data directly, without .hda. Should not be that problematic. The question is why do you need Houdini in this case?
Very cool! Will you try to investigate with USD other Alembic Next Guillaume?
Super stuff
Great work Guillaume! That was something on my to do list, you’re pushing this in the right direction
Insightful! Thanks for sharing 👍
Houdini Teacher & Procedural Pipeline Consultant at E-Houdini Academy
4moI've done some research into this myself recently. I set up a .ABC import method but this can only be imported into the PCG graph with a file path to a file on disk. I also tried the city samples .PBC files, but PCG can't read those... So the method I've so far settled on has been Unreal Datatables. Storing point positions and other data as table entries. These are imported to Unreal using the Houdini Engine, or by importing a bunch of .csv files manually. ( Though manually, I noticed Unreal tends to fail to import them all correctly at times, so Houdini has been more reliable for this ). I've split my instances into seperate point clouds, roughly 50k to 100k per grid. Larger datatable point clouds are slower to save, so don't try to save 2 million points into 1 datatable. The main benefit is that they can be stored in the content browser and shared with other people without them needing the Houdini engine. And PCG can do a range of post processing, such as snapping to terrain, updating the angle, and some culling operations. Though in 5.3 at least, PCG isn't too flexible yet. Its proven more performant than directly from Houdini, as Houdini assets in h19.5 and h20 can be a slow to rebuild compared to loading from cached data.