Blender 2.8 viewport updates

blendertest

(Thinking Polygons) #752

Unfortunately it’s not working here.
This is the build I got:

blender-2.80-32dca36afb1-win64.zip


(darthfurby) #753

I’m using the same build. See if the attached blend file works.customhdritest.blend (1.2 MB)
The following setting needs to be turned on:
sceneworldlighting


(Thinking Polygons) #754

Thanks, your file works and don’t, because there’s a weird bug happening. :slight_smile:

I think I found more or less the bug.
With a fresh scene, if you switch to lookdev mode and then go to the world tab and load your hdr, it will work.
Now, with a fresh scene, if you go straight to rendered mode and then go to the world tab and load your hdr, it will not work.

And more, even if you go the lookdev route to make it work, when you change the hdr in the world tab, it has no effect, the first image gets stuck there.

Not sure if it’s just here, but would be nice if people using Windows could confirm. :v:


(darthfurby) #755

I confirm, same issue on my setup, nice catch!


(Sim_Piko) #757

Speaking of EEVEE and HDRI related issues, is it normal that same HDRI file looks different when loaded with ‘Non-color’ color-space rather than ‘Color’ color-space?
Here’s example with built-in sunset.exr, but it happens on any .exr i’ve tried so far.


(Gtx) #758

Bit worried the beta is almost here but 2.8 is still very unstable. Its still common for it to crash when editing nodes or modelling. While not as problematic as 5 months ago, its still something that needs to be corrected.


(cdog) #759

That should be normal. Loading with non color data basically means using gamma 1 and color data means gamma 2.2. HDRs should be saved in linear space as 32 bit images and therefore loaded with gamma 1. The part I’m not sure of is whether there isn’t an exception in blender’s loading of environment maps.


(ajarosz) #760

In late beta it is normal to crash, not talking about alpha… Beta still means that is still far from being ready.


(razin) #761

the beta will serve as a bug removal so yes it’s normal for now


(cekuhnen) #762

I did not follow all the dev threads so I am curious if baking the indirect lighting is going to be the only option now and not like in older versions being interactive.

particularly with bigger scenes and animation baking frame by frame I can see as an issue

and each time I change a material the color bleeding is off so I have to rebake


(CarlG) #763

I’d like to see the option to bake selected probes only. If you’re layouting the bathroom, you may not want to bother with everything else in the finished apartment, as an example.


(Ace Dragon) #764

Partial lightpath node support in Eevee.
https://lists.blender.org/pipermail/bf-blender-cvs/2018-November/116013.html

There are a couple of limitations, not all of the output types work and the is glossy output does not work with screen-space reflections, but do remember that making everything work with everything is just harder to achieve with rasterization techniques.


(Thinking Polygons) #765

Seems to be fixed already.
https://developer.blender.org/rB11a53ec28a6abd9c0335c6f34b7862003574128b

:v:


(NinthJake) #766

It seems that volumetrics for the World output doesn’t work in the latest Blender 2.8 build (Nov 08 2018)
Can anyone confirm?


(BlackRainbow) #767

I think it would be good to have a something like minimum samples per frame for eevee (and workbench) because during a playback viewport doesn’t have enough time to gather samples and as a consequence shows each frame with lowest possible sample count and with lot of artifacts. is it possible?

Sidenote, just discovered hairsim cache is working in 2.8! great!


(xrg) #768

T/N panels now do a slide animation instead of fade. Gif playback is choppier than it really is.

anim


(Zinogg) #769

that’s futuristic :slight_smile:


(Piotr Adamowicz) #770

I’m getting these weird transparency issues now in Workbench on something that shouldn’t be transparent:


(MACHIN3) #771

Increase your clipping start value,


(Piotr Adamowicz) #772

Wasn’t so much the start value as the end one. Thanks for the tip.

Makes me think the default 1000 is really excessive. Most users won’t be working on kilometer-scale scenes.