SEGI (Fully Dynamic Global Illumination)

Yeah, for all its problems Enlighten realtime GI is good at doing what it's designed to, even works perfectly on mobile VR.

 

Yeah but Beast was removed and only in the final update of 5.x cycle they implemented a similar (yet experimental) solution that still takes forever to bake. Still less forever tan Enlighten. If you're working with BIG maps Enlighten is simply not an option. Beast was fine, removing it was a failure, should have kept and add Enlighten as an option.

 

Seeing as Beast was recently discontinued I don't think it was their own choice completely.

 

Yeah may be but the point is Beast was FAST, and ressults where FINE. I'm not saying to stick with it forever, but the actual experimental path tracing baking solución should have been there from day 0 even ate the actual experimental level. They knew they where shipping unusable lighting solution for many projects as baking for 30 hours each time you make a major change is not very practical. That is why I see a point in SEGI and similar solutions have a super high end profile that takes some secs or minutes to compute best lighting possible wihing that time range and bake it to lightmaps.

 

Is there somewhere I can download SEGI? Even if an older version - I need to test it for our project ASAP to see if it will be a good option for us.

 

SEGI is already a cool solution for my game and I don't need any more options just please don't depreciate it. It worth buying just make this compatible with Unity versions and that would be enough for us.

Why would I need new features when It works good for me already?
Keep SEGI on it's current condition. Only work on a better performance and that would be enough.

 

What already works good for you might not work good for others... I agree on the performance, though. That's something which really needs to be adressed.

 

It is already Depreciated

 

No github link already?

 

What's wrong with the plan. You will still have access to SEGI but will now get the source code too. Sonic has clearly said that he will keep it up to date with Unity and has also said he won't necessarily stop working on developing it further. Basically the only change for your project is you now have access to the source code of SEGI.

I'm wondering if some people are thinking that SEGI being $80 gave our projects some kind of heightened status for using it. It really doesn't. You are not losing anything by other people getting it for free. If your project relies purely on SEGI to stick out then it isn't going to work. You can differentiate your project by finding ways to use SEGI in the most efficient way possible or to make SEGI look as good as possible.

 

I bought SEGI some time ago and now i see it is deprecated !
As a customer shouldn't i still be able to download it... , or there is another place that we will be able to download it from !? I probably missed this on the thread !

 

You can still download it from in Unity. Just will not be updated there so eventually that copy will be broken and you will have to use the git hub version. Is fine now though.

 

Thanks. It is just if i write "SEGI" in search bar in the asset store from within Unity or from the web browser it does not show up !

 

It's gone hombre, will have to wait for the GitHub link to access SEGI from now on.

 

And did you believe that? He makes SEGI deprecated because it sells good and now he want to leave, this happened to all assets he ever created not only SEGI.

I will never buy any asset from asset store anymore because I waste a lot of money on it.

 

It's still in your download area.

 

Let's just WAIT for once and be grateful that he'll open source it? I'm not trying to defend him and I can't speak for his other Assets, but he deprecated SEGI and wants to open source it, because he simply reached the end of his abilities on this. Realtime GI in general being a complex topic, it's understandable. However, this opportunity (making it open source) could benefit everyone in the end. People with highter knowledge can contribute to the project, thus improving SEGI further.

 

I will be surprised if anyone makes a substantial contribution to the SEGI repo. I'm not insisting that it could never happen, but I imagine that code is very complex.

 

Like Mauri said it is still in the download area if you have purchased it. It will not be removed from there unless you manually remove it. In that little download looking box on the top right of the Asset Store web page and the Asset Store tab within Unity. Under S you should have SEGI (depreciated).

 

Complexity is not necessarily a barrier, if the right people with the right skills are motivated. The one of the joys of opensourcing stuff, you just never know and there are often surprises.

Yes the complexity of making a realtime GI system will rule out a lot of people with inappropriate skills, but Im sure some within the remaining group will have a look. And then it comes down to whether the system is well structured and written in a way that doesnt put them off. And a license that doesnt put them off.

I also dont rule out people with more limited skillsets being able to hack some functionality in or out of the system - my coding skills are often on this level and I'm always amazed how, in limited amounts and when realistic goals are set, I can 'punch above my weight' and fix or modify systems that I really have no business understanding and tinkering with.

 

Sonic, great job with SEGI.
You can be very proud man.

Waiting the GitHub link!

 

Can't wait for it on Github. Pray that somebody (perhaps even Unity???!!!) picks it up and continues. OMG that would be so amazing.

 

First of all, I am also a little disappointed regarding the not so great update mentality and then the sudden cancelation of the whole SEGI project... sure, its develepment will hopefully go on under an acceptable licence (personally i prefer the MIT licence), but the fast deprication is kinda harsh and was completed before any real disscusion with your customers... despite you have expressly desired feedback. Thats couldve worked better from your side!

Nevertheless, SEGI is awesome and you did great so far. Its sad, that you lost your confidence, which you really don't have any reason to do so. I believe that SEGI will greatly benefit from new Hardware like the VEGA Architecture. Does someone have som experience with SEGI running on these Cards? They scale at least on paper great througout FP32-16 and so on, which could result in a great performance boost, especially in a system like SEGI, which makes SEGI looking incredible futureproof to me.

And i kinda liked the Idea keeping the project on the AssetStore and including the Sourcecode, that should have been most beneficial for everyone.

So lets see what the future is holding up for this project, looking somewhat excited for the GitHub release and what comes with it. Keep going and and thanks for your great projects.

 

If you want to you can support my official feedback to Unity about SEGI:

https://feedback.unity3d.com/suggestions/unity-should-take-care-of-segi

Please give it a few votes to make it heard by Unity.

 

I'd like to point out that to my knowledge SonicEther has not yet open sourced it, and only stated some options about what he might do.

 

Let's be honest here: There are talented people working at Unity Technologies - take Keijiro as an example. They could have developed something like SEGI or even integrating Nvidia's VXGI in no time. Yet they decided not to do it, probably for the reason I said days ago: broader hardware compatibility. Voxel-based GI won't run on - for example - mobile devices at 30 fps. Even on PC it would still require a decent graphics card to run at 30 fps.

Maybe SEGI needs a sort of baking option, but imo that breaks the initial premise: having true realtime GI without any baking at all.

 

project this complexity will collect dust in github no one will waste large amount of time to understand how it work and even more time trying to improve it for free.
the only option is if sonic work on segi himself in gethub
that's a sad thing but it's realty.

 

lol finally someone says it

 

I have a newbie question:
I've played the SEGI demo, and it was the first time I felt something was really photorealistic.
Can somebody explain what it is that causes it?
There's only 1 light in the SEGI demo, like in real life where there's only the sun, right?
This means that there's no light faking, everything is calculated as in real world, and that's why it looks so real. Also, developers don't have to spend time on faking light (which looks unrealistic if not done absolutely perfectly, which is an art of itself).
Is my explanation correct?

 

As compared to what?

 

In the SEGI demo (the hallway) there is a quality setting. When I put it to 1, I get the feel that I get when I watch any other demo scenes (sorry, I don't remember any, it was just a feeling that I gathered during the last weeks of lurking. Some scene were also YouTube videos). Quality 4 blew me away, but it raised the question if I perhaps just don't understand Segi and there's something else that I should pay attention to when watching this asset in action.

 

SEGI does GI. GI means you have light bouncing around (as it happens in real life). The advantage of SEGI is that it does it in realtime, the disadvantage that it's expensive.

Enlighten also does GI, Progressive Lightmapper also does GI. The discussion about light faking is more or less irrelevant.

 

And the "Quality 4" in the Sponzo demo of SEGI can be achieved with Enlighten as well? All the "cone" and "voxel" things?
Also: Why are you so interested in SEGI?

 

Yet another deprecated beta. Seriously!

 

Sure? Depends on what you mean. It would look "different", but technically equivalent (you would have precompute GI from enlighten, reflections from cubemaps and SSR).

Is that towards me? Did I express intense interest in SEGI?

 

@AcidArrow Yes, that was meant towards you.
Did you actually mean "Sure?" or "Sure."?

I understood your engagement in this forum in such a way that you were interested in SEGI.

 

Is this on GitHub now?

 

I own SEGI, I played with it for a few days. I had fun. I'm not using it for production.

I did mean "Sure?". Because "Quality" is subjective so I can't answer definitevely.

 

Enlighten doesn't use "cone tracing" and/or "voxels". Here's a blog post from Geomerics (the
developers of Enlighten) explaining how their system works:
http://www.geomerics.com/blogs/demystifying-the-enlighten-precompute/

 

I would so much like to see the SEGI hallway with Unity-only effects done by somebody who's experienced with Unity. Because I really can't compare SEGI to anything else with my knowledge. Right now the SEGI demo just reminds me of an asset that I want to own because the demo looks like the asset can do something that's not possible otherwise. While perhaps it is.
And this is also one of the main reasons which so many people buy assets at all, I think.

 

It is absolutely possible to achieve near photorealism in Unity without SEGI. It is also possible in UE4. You don't need SEGI to achieve high quality visuals.

 

Since segi I have start to educate myself about other GI and lighting system, the only difference is "scope", offline solution tries to solve an entire scene with high sampling rates and fine data (that mean all pixel of the scene) while real time solution only solve the immediate surrounding of the camera up to a limited range with low sampling rate and coarser data like voxels (with variation of that to increase range, aka cascade).

For example enlighten segment the scene into "surfaces ID" and just maintain a precomputed weighted visibility list limited to the n biggest contributor, then after each surfaces is lighted, it compute the weighted sum of the weighted contribution of each surfaces in the list of a given surface. Conceptually voxel solution do the same but the visibility is computed in real time at a point in space instead of real time. Both limit the contribution to static geometry in the face of varying lighting condition.

It seem the main problem is all about the visibility calculation that tend to cost the more.

 

What stroke me about the Sponza demo was that it was driven by one light only.
I always felt it was wrong to introduce several lights into a living room that should receive its light from outside.
Can anybody shed some light (haha!) onto this?

 

What question do you have, specifically? I know the interior archviz scene posted on the SEGI website is using emissive quads to shine light into the room, rather than a directional light; maybe both, would have to ask Cody.

 

You keep asking very similar questions. GI does bounce lighting, so 1 light can illuminate a lot more of a scene than it otherwise could using only direct lighting. Segi does this, Enlighten does this. Using a single directional light + GI will indirectly illuminate the whole Sponza corridor. Theoretically Enlighten will look better.

The advantage of segi is you could have procedurally generated the Sponza scene and you would still see bounced light. You could also have flaming torches move through the scene and they would properly contribute to GI.

Enlighten can look fab, it's mostly just the workflow and wait times that make it a pain to use, but once it's precomputed it's extremely fast.

Also we shouldn't overlook the importance of quality tonemapping and post which a big chunk of the community seem to overlook.

 

People do that because:

1. It's a relic from the time we didn't have GI.

2. They're going for more filmic photorealism and films introduce several fake lights into a living room as well.

 

Hi Guys,

I am developing interior/exterior arch. design tool and the whole scene is dynamic. Unfortunately the interior lighting is missing indirect lighting and I found SEGI would be an awesome solution. I hope the github will be available soon so I can try that out.

I believe personally that there will be VXGI solution for unity sooner or later since Unreal engine implemented it recently.

The project is WebGL 2.0 based .. I know that SEGI is designed for Direct X 11 .. but as long as you don't need tessellation or compute shaders you can do almost anything .. so I believe I could potentially maybe modify SEGI to work on WebGL 2.0.

I cannot wait to try it out. Is there a way to download some older version from somewhere ?

 

I can´t wait for the update with blue noise and cascades I think its gonna be mindblowing as it is.

 

Hi Sonicether,

I know how you feel I have been chasing my dreams like you too and ended up giving up after two years of full-time work to realize that the project is just way too much for me to handle. It was a big lesson. It's funny but I used to be a huge fan of real-time GI back in the day and also develop myself a semi-precomputed GI solution that works with dynamic lighting something similar to Enlighten but not that advanced and optimized .. I gave up on this too .. but it was fun and it worked !. I do still code a lot of GPU stuff, so maybe If I will be able to use SEGI in my work project I could maybe add my contribution to the project and help with optimizations. I wrote a GPU based PBR bidirectional path-tracer in OpenCL with all basic rendering features for my app back in the day so I really love rendering. However all this work is incredibly challenging and time-consuming, it's also hard to make money with this alone if you are into business ..
I wish you best luck with your dream chasing and I hope to see SEGI on GitHub soon.

Andrej

 

MIT is ideal, although does require developers to voluntarily contribute back their changes. It is a vastly simpler license however, and much easier for companies to adopt. If you're open to suggestion - it's the one I'd go behind (and have done for my own open source contributions over the years...)

I would also raise a concern, that GPLv3 is likely incompatible with open source projects built around Unity - as Unity is linked to the code, and Unity is not open source under the GPLv3; this would make redistribution of the open source SEGI legally impossible by anyone but yourself.

That said; once this is open source - I'd be happy to help sponsor development on the open source SEGI with at least a little bit of cash - it does fit a hole in Unity I would love solved, and if that helps things move along in it's post-commercial life, all the better.