Thanks for sharing! I was there in the audience at the time, but haven't rewatched it since. I'd really love to see more of these talks on how a solution was achieved in addition to how the solution works.

Always fun to debug these types of issues where your prepass for example uses an FMA somewhere for determining vertex positions while your main shading pass somehow doesn't!

I think you might've convinced me to give this a proper shot. I was never against the idea of using LLMs to assist in writing code, but for some reason it never seemed to register as something I'd inherently want to use. It is appealing to be able to get to the interesting hard problems more quickly

On Bluesky chat! Basically just messaged that we're around at that time!

DM'd you!

We didn't! Just passed by! Open to an invite though, haha!

You were major part of making this all happen! It's good to be able to celebrate all that hard work.

Congratulations Trung!!

Do you end up storing UVs in your gbuffer at all in that case?

I think I remember you mentioning in one of your talks that you have a lean gbuffer by not writing color to it and instead sampling colors based on material ID. Does VT allow you to support textures this way as well by guaranteeing you'll only ever have one texture atlas for all materials?