Ben Golus

@bgolus

Tech Artist, Graphics Programmer, and general Game Developer. Current project:

Kirkland, WA
Vrijeme pridruživanja: kolovoz 2010.

Tweetovi

Blokirali ste korisnika/cu @bgolus

Jeste li sigurni da želite vidjeti te tweetove? Time nećete deblokirati korisnika/cu @bgolus

  1. prije 17 sati

    Ever wonder why baked normal maps just don't look quite right in Unity? I did, so I decided to find out why.

    Poništi
  2. 2. velj

    My deep dive into tangent space inconsistencies between applications and Unity has led to a curious discovery. Unity uses per vertex binormal recalculation, _except_ the new VFX Graph which appears to use per pixel. 🤦‍♂️

    Poništi
  3. 2. velj

    Except when they absolutely should be, like for MikkTSpace tangent basis, or depth projections, or parallax offset / occlusion mapping ...

    Poništi
  4. 2. velj

    No game is ever truly finish. Every released game came out because the devs got tired of working on it or ran out of money.

    Poništi
  5. proslijedio/la je Tweet
    2. velj

    A small prototype: I generated a tile-able liquid-mesh with and move several instances of it along an spline in . More details in my -thread:

    Poništi
  6. 1. velj

    As far as I can tell means only xNormal and Substance can bake correct normal maps for Unity and . Every other tool that uses MikkTSpace uses per pixel binormals, like UE4. Everything online seems to suggest UE4 was the odd one here, and it turns out not to be so.

    Prikaži ovu nit
    Poništi
  7. 1. velj

    Went through the source code as far back as I could. Looks like Blender really has always been calculating the binormal per pixel, which means it really has never matched Unity's MikkTSpace implementation.

    Prikaži ovu nit
    Poništi
  8. 1. velj

    Does anyone know if changed their tangent space binormal calculations at some point? Lots of places say it matches Unity 5.3+, but at least for Blender 2.8 that's wrong. Blender 2.8 is baking using per pixel binormals, like Unreal.

    Prikaži ovu nit
    Poništi
  9. proslijedio/la je Tweet
    29. sij

    Woohoo! After 6 years, The Pedestrian is finally out. Thank you all for the support over these years!

    Poništi
  10. 29. sij

    Oh, and this isn’t getting vertex positions still, just world space distance to edge. Should be able to do a few cross products & dot products to extrapolate the vertex positions from what’s there.

    Prikaži ovu nit
    Poništi
  11. 29. sij

    The issues with it going bonkers when perspective is in the mix is kind of expected, but it also goes catawampus with orthographic views when viewed off-axis, which makes less sense. This is the kind of minutia in understanding of derivatives my brain hasn’t quite grokked.

    Prikaži ovu nit
    Poništi
  12. 29. sij

    Sort of solved it on my own, at least for the use case I wanted to test it with, and only when the triangle is aligned to the view plane.

    Prikaži ovu nit
    Poništi
  13. 29. sij

    Graphics programmers / shader wizards out there. Somewhere I saw someone post an article on deriving the vertex positions of a triangle in the fragment shader using barycentrics and pixel derivatives. I can't for the life of me find it, anyone else remember it?

    Prikaži ovu nit
    Poništi
  14. 27. sij

    I should do another pass on it to center the grey better.

    Prikaži ovu nit
    Poništi
  15. 27. sij
    Prikaži ovu nit
    Poništi
  16. 24. sij

    Anyone know what this illusion is called? For those wonder, "what illusion", there's no dark or light vertical edges in this image. It's just a saw toothed gradient. I'm going to call it Mock Bands, because I'm pretty sure it's not Mach Bands.

    Poništi
  17. proslijedio/la je Tweet
    23. sij

    Twitter wisdom! In OpenGL, should compressed *3D* textures with S3TC (aka DXT) formats work? Can't quite get them to work (works in D3D/Metal/Vulkan fine). The S3TC spec seems to imply they should not work (), but no such errors as in spec.

    Poništi
  18. proslijedio/la je Tweet
    22. sij

    this is one of the damn geekiest things i have done in , just made a tool for to wrap ropes around any mesh.

    Poništi
  19. proslijedio/la je Tweet
    20. sij

    Wrote a post about normal reconstruction, described a method based on 's SDAA idea, which is more costly but also more accurate than 's method.

    Poništi
  20. proslijedio/la je Tweet

    Fellow Men. We need to talk about the way women are (mis) treated in this country and in this world. Here’s my thread. /1

    Poništi

Čini se da učitavanje traje već neko vrijeme.

Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.

    Možda bi vam se svidjelo i ovo:

    ·