why sloppy?
02.10.2025 15:09 โ ๐ 0 ๐ 0 ๐ฌ 1 ๐ 0@ghislaingir.bsky.social
UE content creator, specialized in tech art. I don't share my opinion much.
why sloppy?
02.10.2025 15:09 โ ๐ 0 ๐ 0 ๐ฌ 1 ๐ 0I'm very confused about the Render Heightmap node. It seems to be using two channels to store the integer and fractional part of the world height but I've no idea where the range come from. It doesn't seem to be normalized based on the landscape Z bounds.
02.10.2025 15:08 โ ๐ 0 ๐ 0 ๐ฌ 0 ๐ 0checking the overlap is what confuses me. For a sphere I could get around, but for an arbitrary mesh which voxelization could suffer from severe aliasing issue because of low resolution, I don't see how I can reliably detect if a voxel is inside the mesh
02.10.2025 10:19 โ ๐ 0 ๐ 0 ๐ฌ 0 ๐ 0Hacky experiment. Tried to generate a distance field of an animated skeletal mesh using separable JFA. Performs well but I don't know how to make it signed/handle voxels inside the mesh ๐ซ
01.10.2025 14:55 โ ๐ 9 ๐ 0 ๐ฌ 2 ๐ 0Another sdf-related tool. Nothing too fancy, this one bakes a 2D distance field for actors containing spline component(s). Might come in handy for shorelines etc.
I'll do a breakdown of the scriptable tool BP in my upcoming video.
Every time I try to generate a material via BP/Python, I hit dead ends๐ฉ. Was hoping to generate a volumetric material asset, but apparently, you can't connect anything to the Extinction material property...
Doing that in CPP isn't worth investing the time for my particular case.
A small tool I built the last two days. I'm using Geometry Scripting + Python to generate SDFs from any geometry in the scene and visualize the result.
This is part of my next video dedicated to SDFs
Currently reassessing my plans for 2026 and whatโs next. Lots to think about. Struggling with mental health a bit lately, and Iโve been feeling a dip in creativity and motivation.
22.09.2025 13:08 โ ๐ 5 ๐ 0 ๐ฌ 0 ๐ 0Rough locomotion tests. It felt cute at the time.
22.09.2025 13:08 โ ๐ 5 ๐ 0 ๐ฌ 1 ๐ 0Character I prototyped last year. I had the idea for a small party game, but I could never quite find a healthy balance between personal and professional projects. My brain doesnโt let me do 50/50.
22.09.2025 13:08 โ ๐ 10 ๐ 0 ๐ฌ 1 ๐ 0Yeah each use case has its own workaround but still, DF are great and can (could) be used for so many things. Sad to see it's no longer an option.
15.09.2025 09:18 โ ๐ 1 ๐ 0 ๐ฌ 0 ๐ 0I wasn't aware of the multi-sampling trick. Far from ideal indeed, but good to know, thanks.
I once stumbled upon that ticket but it doesn't say much... I'd love to see an official word about the technical reasons DF are broken and why it likely won't be fixed.
What's the official statement on using distance fields in shaders? Afaik, they've been somewhat broken since 5.0. Workaround is to use a CVar to only use one low resolution cascade, I think, but it's far from ideal. Am I wrong? Any clues? Ty
14.09.2025 13:00 โ ๐ 3 ๐ 0 ๐ฌ 2 ๐ 0The paper: mmacklin.com/pbf_sig_prep...
11.09.2025 23:33 โ ๐ 0 ๐ 0 ๐ฌ 0 ๐ 0First time trying out the rasterization grid interface as well. Seems pretty straightforward. I'll try to make a tutorial if I eventually manage to make sense of what I'm doing (:
11.09.2025 23:33 โ ๐ 6 ๐ 0 ๐ฌ 1 ๐ 0Decided to try Position-Based Fluids for a first go at realtime fluids. I think I got the algorithm wrong, I've got weird boundary behavior. Meh, I suck at maths, it's an infinite pain to go through academic papers :(
11.09.2025 23:33 โ ๐ 11 ๐ 3 ๐ฌ 1 ๐ 0TIL that NDC has a new method for querying particle data in BP :) The basic callback interface was severely limited. This looks much more powerful. Neat!
08.09.2025 15:48 โ ๐ 6 ๐ 2 ๐ฌ 1 ๐ 0Continuing with my Niagara/Grass experiments... Here, a regular grass clump mesh is instanced about 64 times and each card is injected into a buffer to be individually simulated. I still need to fix plenty of issues & write to velocity buffer to fix TAA artifacts but it's a start
01.09.2025 23:47 โ ๐ 14 ๐ 1 ๐ฌ 0 ๐ 0Core C-Side is my last golden berry, I swear. Golden B Sides and Farewell are on a whole other level, I want to keep what's left of my sanity. Still, fairly happy with my progression :shrug:
28.08.2025 23:17 โ ๐ 0 ๐ 0 ๐ฌ 0 ๐ 0Files are available as a Tier 2 reward on my Patreon:
patreon.com/c/GhislainGir
Thanks for the on-going support ๐ฅฐ
Onto the next!
Cheers
๐ฅณ๐ฅ New YouTube video just dropped!
A quick 20-min dive into the basics of Niagaraโs Dynamic Mesh interface + a breakdown of a more advanced (but flawed/experimental) use case.
Link โฌ๏ธ
#UnrealEngine #Niagara #VFX
In the meantime you can find plenty of cool projects like this on my Patreon: patreon.com/c/GhislainGir
And plenty of tech deep dives on my YT:
youtube.com/@ghislaingir...
Have a great week end โ๏ธI'll be on my bike all week end long. Hope you get to do something cool. Cheers.
Anyway, time to clean the project up and document it.
I'll release a video explaining how this is done next week and I'll let you be the judge of what to make of this :)
It seems highly experimental still (no shadow casting?) and some aspect of it is CPU driven (bounds, vert & tri allocation, etc.) so I'm not sure if it's viable to do Ghost-of-Tsushima-like grass? Out of my league.
22.08.2025 22:46 โ ๐ 0 ๐ 0 ๐ฌ 1 ๐ 0Alright, I'm done with my deep dive into Niagara's Dynamic Mesh interface. LOTS can be done with it.
Frustrum culling is aggressive here for demonstration purposes.
Overhead seems to be high: 1.3ms for generating this :( I'm sure my code isn't as optimized as could be tho.
#UE5 #vfx #gamedev #indiedev๐ฅณI just released a short video explaining how sphere tracing works and how to write a sphere tracing function in HLSL for collision detection in Niagara:
www.youtube.com/watch?v=B4RR...
I have seen all his videos, yeah :D I haven't poked him, no. Maybe I will :)
21.08.2025 09:24 โ ๐ 1 ๐ 0 ๐ฌ 0 ๐ 0Only way I can think of is using Niagara's Data Export interface but I can hardly see how you can hijack it to export the necessary data for collision generation :/ I'd say, no, not out of the box?
20.08.2025 23:03 โ ๐ 1 ๐ 0 ๐ฌ 0 ๐ 0It's not just about tools, it's a totally different way of rendering grass that opens the door to lots of possibilities. It's not without downsides though.
www.youtube.com/watch?v=Ibe1...
It's just an experiment and the first thing that came to my mind to test Niagara's Dynamic Mesh interface :)