I’m also working in a NPR thing… and although the EEVEE-Next seems indeed different from the Eevee-Legacy, in my case still seems to look better than before, even as in NPR.
Therefore it keeps the NPR look and improves the definition of it thanks to the new shadings… for me that is working in my favor
The Agx Color space is actually giving me more troubles than the new shading in EEVEE But I’ve also learned to overcome it and found solutions.
I’ve been following the developer notes for a long time.
eevee next required a higher version of openGL.
But Mac is not supported, so the developers had to spend a lot of time on METAL support.
All the while, the time for eevee next’s debut was getting closer and closer.
Naturally, the development period was too short.
Developers are in a very difficult environment where they have to satisfy the immediate needs of users and also have a future base for Vulkan.
Of course, this may be a personal opinion.
Whatever the reason, my concern is that developers will burn out.
It’s amazing what they’ve been able to accomplish in such a short amount of time.
I have been using Maya for over 15 years in the industry and have moved on to Blender.
The reason is eevee.
I am a huge fan of eevee and really enjoy it.
It will probably continue with eevee next.
Rendering might be slower now, but improvements on that front are likely to come after the initial release, right ? that’s how things usually go. And the quality bump is frankly impressive… It looks very close to Cycles.
That’s what I’ve been saying! we should all be VERY VERY thankful of the developers and their constant constant work! And again the software is all free! And I think it’s totally fine if the developers release the engine in a “bad” state, because again, every render engine and it’s first launch is always not good. And it takes time to improve it! Eevee legacy when it came out, people though it was great and at the same time terrible and over time it improved!
And right now the blender developers are in a constant burn of emotions, they are really tired. It’s hard work! And I feel that as a developer myself. So please, lets all be patient with Eevee NEXT. I am 100% certain it’ll turn out amazing if we give it more time!
I am very thankful to the developers! Thank you very much!
We are just sharing our experiences, for sure I don’t want to cause any pressure to the developers.
It’s good keep posting the things we stumble on and talk about them… no point in holding things back Talking freely about the good and the bad usually improves things and that was/is the magic of forums!
If EEVEE Next is not sufficiently stable in time for Blender 4.1, we may move it to 4.2. We’re trying to get it ready for 4.1 and are working under that assumption, but it’s not a decision set in stone.
I have the strange impression the build bot not compile the daily builds but the last day daily builds let me explain if we are the 20 and he do release the files of the 20 he build the files from the 18… With one day decay… I see that when jeroen are commited the mid gray preview ball HDRI lookdev… Is very strange !
I tend to interpret things far more literally at times, than someone intended… and there are moments, such as this, that what should be “logical” just lays on the side of the road and not considered.
It’s fine if the NPR workflow moves to a gbuffer approach with the real time compositor. It’s a more flexible approach and you can do stuff in real time that wasn’t possible before.
But for that to happen, we need:
. Feature completeness in the real time compositor, including arbitrary cameras, view layers, passes, data, etc. Not even the Z pass works right now
. A way to make the multi sampling antialias at the end of the composition. The postprocessed antialias node we have now is not sufficient. It would need to be actual jittered sampling of the 3D scene. But right now, in both eevee and eevee next, most sampling is tied together regardless if it’s for shadows, reflections, textures or antialias. As an user maybe I’d want to sample shadows and reflections first so they’re smooth, then apply a colorramp cell shading, and then sample again to antialias that cell shading.
I’m OK with Eevee next being delayed to 5.0 if it means to have a complete real time compositor and a new NPR pipeline that works out of the box, without the need for gooengine
thats not entirely true , there are many things which could be “frozen” in animation playback and rendering as demonstrated by other softwares (even blender game engine), and unfrozen afterwards.
then there are more creative approaches like maya using gpu skinning, gpu tesselation hardware and alembic gpu playback for realtime viewport optimization (and conversions too)
maybe its time too move eevee legacy to an addon like cycles and other rendering engines are.
It was always feature constrained by the blender foundations design goals and then we are able to add the missing stuff and nodes from gooengine, malt and other stuff we find useful.
Eevee is an awesome engine which does alot of stuff right, but sometimes i wish it just had more features like lineart shaders based on nomral and depth pass or other non intrusive postpro shaders like those gameinjectors(and even some videoplayers) have.
One thing funny in Eevee-Legacy is that the GPU, on the creatives I make, is most of the time only at 25%, this is so much true that I can lunch the scene 4 times (using this add-on) and all of them render at the same speed
On most cases (not true when the scene as many volumetrics) as long the card has memory will just run another instance as if it wasn’t doing a thing.
On one creative, that used about 1.2Gb of ram and contained no heavy effects or shaders, I lunched 8 instances and it rendered about 6 times faster than just one… so undoubtedly that the GPU has room to do more
Can you just start reporting this on bugtracker?
Every day there is a new build, what is the purpose of notifying every person in this thread that it doesn’t work?