SEGI (Fully Dynamic Global Illumination)

Games that have levels that are procedurally generated or modifiable by the players CAN'T bake, like I said. For example, in my game, the levels and terrain are all randomly generated, and the players build structures out of blocks and place objects everywhere. There's literally nothing to bake in the editor.

 

This is obvious
What should I accept?
I have already accepted this because this is obvious

I just make comparisons to help people don't lost performance if they don't have fully dynamic worlds

 

Then why do you keep talking about performance comparisons for static scenes where you can bake the lighting ahead of time? That's a use case that SEGI was not designed for, so stop comparing those numbers.

Conversely, you can't even draw comparisons of SEGI vs Enlighten for dynamic scene geometry. Enlighten can't be used for that use case.

 

If it's obvious then you shouldn't keep saying things like Enlighten is for REAL developers and SEGI is just for beta testers. Baked GI like Enlighten is for games where the game world is mostly static and unchanging; Realtime GI like SEGI is for games where the game world is dynamic and modifiable. It has nothing to do with which one is more "real" or "AAA" or anything at all like that.

 

OK OK OK

 

In this case, i told that the GI is not very important to make the game real. Post processing is more effective in action.
The more important part is the indirect shadows that we can simulate it using AO
So :
Indirect shadows using AO
GI using Light probes

The result is same

If you are a expert user you should understand that what i'm talking about(Frostbite games in action)

 

I understand that. I even agree with you. GI can be "faked" sufficiently enough with good post-processing effects like AO, color bleed, and auto-exposure. It still doesn't discount the fact that you are slamming SEGI's performance for a use case for which Enlighten is optimized and SEGI was not meant to be the ideal answer for.

No one thinks SEGI performs better than Enlighten for static geometry. No one should just think SEGI is a bad product because of that either. Use the tool that is best for your use case.

 

Okay but then we need a debriefing:

Someone else (@ShadowK made comparison across many techniques and platforms and didn't attract similar debates:
- he use the same scene with the same assets,
- he composed the scene himself using similar set up with light and camera,
- he tweak each techniques and platform himself, which mean the skills is consistent across all scenes

This reduce uncertainty a lot, the only ambiguity is workflow vs technical implementation, which, as a dev, is the same as it tells you how fast a technique allow to reach a given visual target with a similar skillset. That's perfect, even though someone else would obtain bigger results with a given technique, it would be a function of investment in that technique, for a comparison what ShadowK did is great because it flatten concern. Of course the discussion then evolve naturally to other concerns and use case (what about open field?).

Meanwhile your comparison is from a scene and assets, composed by someone else, with a different skill set, on another platform, nothing is stable in that comparison because too many unknown. You tried using the same asset but the import isn't perfect, for example the computer below the table is white on unity's version but have a black facade in the UE's version, that doesnt help as the worry about uncertainty is already here due to the point I had outline previously.

As the image lack important fundamental images quality, that you seem to dismiss, that doesn't help to pin point what's comparable, is it the translation from skill set? is it the different artistic sensibility, the workflow, the botched asset import, the raw technique, the algorithm? We can't know, therefore it attract negativity. It doesn't help you kind of fueled the fire with bold claim and starting calling people out!

 

Why bake when you can microwave...?

 

**This is an off-topic post**

In light of the UE4 vs. Unity talk, I wanted to share: This is a scene converted by Jeryce Dianingana for Unity 5, originally published for UE4 by Clinton Crumpler. Jeryce has composed some of the highest quality Unity environments I have seen, and it's interesting to see how a pro approaches the challenge of replicating the look of a UE4 scene in Unity using the current feature set.

 

I'm not sure it was unity vs unreal as much as the methodology of comparing different solutions (aka segi vs unreal vs enlighten vs post process).

But this scene as the same problem than the @aliyeredon2 scene, there too much interpretation going on to be a good comparison, it just tells you can have good result™ but not compare the solutions. I mean asset are downright different, with different scale and positioning. And I don't like differing camera angle because view dependent effects can change how you perceive a scene.





The image above look more realist but it might not be the lighting as much as the more sensible scale of objects, it also have a better composition, so what give about the technical difference I see when the interpretation mess so much with the perception? How do I know it's technical vs artistic judgement? I do have an opinion but that's not strictly informed and might be wrong. IT doesn't inform the workflow at all.

Another angle





But I'm being nit picky, this is way ahead good "enough" as quality to show whatever platform can do something very interesting in rendering, but I won't sit in that chair on the last image.

 

Finally Downloaded SEGI, Wow! it's awesome!
I would've been happy to pay the asking price.
I've also used some 2k textures and the unity post processing stack.
Any comments? Advice? I could use advice...

 

thank you for your reply. lighting is important , but the effect should be used with SEGI, can you help me to make a interior effect like that

 

hi , can you shere thie project , it is good effect for me, thank you very much

 

it will be the best if you can place some screenshot here

 

Here is a revised version:

I used uber shader for the materials and segi for the lighting.
I'll post my post production and shader settings soon.

 

Simply awesome.

Full album:
https://imgur.com/a/l0qQq

 

A scene I started working on this week using SEGI. Still very WIP, a lot of stuff not modelled or textured properly at this stage

 

This is some proper SEGI use right here, emissive surfaces really working nicely. Nice geo too, love that insulation material on the walls +1. Definitely post the final result when you're done!

 

Thanks man! I shall do

 

Using emissive objects to fake point lights.

Awesome results so far (yeah those round lamps are F***ing bright).

 

@sonicether

Any chance to make it possible to calculate GI for the entire scene, and then export it as a lightmap?

This kinda defeats the purpose since it's not dynamic anymore, but it would be amazing to have a lightmapping tool that only takes a few seconds to calculate.

I know there's the "update GI" toggle, but if it's disabled it won't cover the whole scene, plus it still takes a lot of performance in comparison to having SEGI disabled.

I'd gladly pay a lot for something like that.

 

From 2016

 

HOLY S*** YES

I really hope this will be a thing soon

 

lol, I think I meant to imply that it's been over a year since this was requested - and I don't think that was the first time it was requested either. As such, I'm not sure I would hold my breath for that feature. If you're looking for high quality lightmap baking, you might turn your attention to the new OTOY integration.

 

I'm not looking for ultra high quality, I'm looking for speed.

SEGI could create a lightmap in seconds while Enlighten or PLM would need hours.

 

Check the Otoy backer that is being integrated now. Link in previous posts of this thread.

 

Damn, finally!

Didn't expect Octane to be that fast.

Octane lightmapping isn't out yet, right? Or is it included in one of the Asset Store sample projects?

 

AFAIK is kind of preview mode, its downloadable tho, I've not yet tested. I think it bakes only a single object for now so you've to like merge the mesh. Obviously I asume this is a temporary limitation. You may whant to continue details on otoy stuff on its own thread.

 

Can anyone tell me if there's a way to make SEGI work with Pre-Integrated Skin Shader?

Would be willing to pay someone to do this for me, I have no shader skills whateversoever... If it's possible.

 

Hi everyone

The reflection of SEGI always blur as below image cabinet doors

Even matllic and smothness are set to 1

How could I make it clearly?

Thanks

 

Aren't the SEGI reflections based on the voxelization? The advantage being it's not limited to screenspace geometry, but the tradeoff is you don't get true reflections of the objects either.

 

Yes, you need to use SSR and box projection for accurate reflections, not SEGI. This was something that was discussed in the thread *way* back.

 

Ideally, you should use normal or roughness maps. SEGI's reflections will never be mirror-like, because the reflections are based on the voxelization itself and upping the resolution would impact the performance. sonicether talked about it in one of his videos:

 

Alright, thanks for reply
I'll try another way.

 

**EDIT** I had an older version of SEGI apparently? I thought i was up to date! None the less, problem solved.

Does anyone know why SEGI works on 2017.1.1f1 but not on 2017.1.2f1?? It spits out this error:

SEGI is not supported on the current platform. Check for shader compile errors in SEGI/Resources

Same machine, same project. No changes made apart from migration. But even if I import SEGI into an empty 2017.1.2f1 still the same issue!

 

I'm sure it would be beneficial to post said errors.

 

Haven't seen the github build updated in the last month. Is someone working on it still? Also is Sonic still working on it?

Just curious

 

It's just the debug in the SEGI.cs. I didn't have any other errors or console warnings at all.

 

The last time I created a new pull request was 26 September and it's still open without any commentary. It's obvious to me that the original creator is no longer interested in his creation... So I continued to work on SEGI but in a private fork. I have already made over 400 changes to it, including new baking version that is working lightning fast.

 

Thanks god there's no more dull comparisons and nagging spam anymore... thread got its life and dev talk back. Any words from the developer recently anywhere?

 

Holy s**t, SEGI-based baking is incredible news.

What are your plans or ETA for a release?

 

You can bake normal unity lightmaps from it?

 

@nxrighthere ^echoing this comment, can you describe a little bit about how SEGI baking works in your implementation? Do you have any screenshots you could share demonstrating this functionality?

 

It looks like the plugin is turning in a baking solution.
How fast it is compared to Octane baking ?

 

It would be cool if Baking could be enabled or disabled, or it was a different component all together. What i'd like to see personally is the ability to bake during run time, clear the baked maps and have the ability to rebake scenes. Could prove extremely useful for procedurally generated games and other things (like my project).

 

Progressive lightmapping by caching player's cone of vision sound like a dream lol.

 

Are you planning on making it public?

 

I know you'll probably see us non coders as some kind of beggars XD, but I'd like to ask, have you considered opening a github with your own version of SEGI? You seem to be much more active than Sonic right now, and it would be a bless if everyone could try it and give you feedback!