was my first time tracing over a concept sketch. it came out better than i expected. the cool thing is that it only took a couple hours to get here. before i would try "ideas" directly in blender... i didnt realize how much of a time sink that is even for simple stuff like this.

modeling with only a vague notion vs with a pre-existing 2d reference is like night and day. i already "knew" that rationally, but i've learned that on a visceral level today.

yes i know these things would never fly. shmups have their own laws of physics though.

remebered the melancholy of haruhi suzumiya today for no apparent reason... looked at the date, Aug 18 is the start of endless eight arc. there's no such thing as a coincidence!
https://www.youtube.com/watch?v=WWB01IuMvzA&list=RDWWB01IuMvzA&start_radio=1

i'm slightly proud of this one because it really started life as just a circle (actually, a hexagon) and i didn't follow any particular tutorial, just tried to apply whatever stuff i had learned earlier.

maybe all these blender tutorials are finally starting to pay off

I will be at siggraph in LA this year, hmu if you are also there and would like to grab a drink/coffee/lunch/whatever

I want to go to here; this is the future they stole from us.

relevant bit that defines the parameter of trowbridge-reitz distribution as the "squashedness" of the ellipsoid ("e" is basically alpha here, replace sin^2 with 1-cos^2 and the usual form will become apparent)

alpha controls how squished the ellipsoid is: zero alpha = no curvature at all (degenerates into flat disc) so all micro-normals are oriented the same.

in the anisotropic case, as far as i understand, the two alphas are slightly different: they control how much the ellipsoid is stretched in the tangent/bitangent directions.

somehow it had never occurred to me before - most practical PBR guides describe "roughness" in a rather handwavy way, so you get used to thinking about it as just some number, but it actually has a very precise meaning.

trowbridge-reitz/ggx models the distribution of micro-normals within a surface patch as if the patch was part of a perfectly reflective ellipsoid.

I did Compiler Explorer's monthly accounts, and for the first time in ages we only broke even. Usually we run at a small surplus (which I save) & we have a year contingency saved up.

If you want to support us:

* Buy stuff: https://shop.compiler-explorer.com/
* Patron/GH: "Other" menu on CE

Would appreciate a repost - thanks so much! :)

can someone who understands this stuff explain to me why crowdsourced ml training is not a thing? afaiu when training you _already_ have to use many gpus across many machines potentially in different data centers. would the same techniques not work for distributing the training process across many "peer" machines?

22 years later, I made a "remaster" of our old demo Syntonic Dentiforms. Now on Windows, Linux, Mac and Web, with cleaned up source code and ported from ancient closed source libraries (D3DX, etc.) to sokol_gfx, sokol_audio, stb_vorbis (they are nice!). Here's a blog post: https://aras-p.info/blog/2026/04/13/Syntonic-Dentiforms-redux/ :demoscene: #demoscene

INTRODUCTION TO SPHERICAL HARMONICS FOR GRAPHICS PROGRAMMERS

finally done

https://gpfault.net/posts/sph.html

is spherical convolution (as in, convolving functions on a sphere) really such a niche operation? the only definitions i can find are all 10 pages deep in some ml papers OR on someone's lecture slides from 2015. you'd think the wiki article on convolution would at least mention something...

an oddly satisfying illustration why the number of spherical harmonic functions up to and including a certain degree is always a square.