SEGI (Fully Dynamic Global Illumination)

I haven't had any changes personally.

 

Im using SEGI in VR projects. It works perfectly fine (besides the low performance of course)

 

found the issue for slow performance after unity update
on my laptop there is 2 GPU intel and GTX980
after unity update it ran on the integrated GPU by default.

 

Hey everyone! The new update is nearly done. I just need to merge the cascaded and the non-cascaded versions, update the demos, and update the User Guide. I wanted to discuss a few things with you guys, however, regarding the future of this project. These things I have discussed before in this thread, but I want to involve you guys in the important decisions that I feel need to be made.

So, as some of you heard from me before, when I released the first beta version of SEGI, I had a lot of ideas for how the project would improve and evolve, but I didn't expect nearly none of those ideas to work. The development process of SEGI has been a grueling process of trial and error (and error, and error). The new update with cascades aside, I've come to realize that I unknowingly released SEGI in a state that was essentially at the peak of my abilities to create.

SEGI is nestled in a valley where I feel like any step I make in one direction breaks the practicality of the system in the opposite direction. For example, implementing directional voxel data, to the best of my abilities, ended up being too slow to be practical. Implementing volume-traced (instead of screen-space traced) GI to speed up tracing made the visual result less robust and therefore impractical (even more light bleeding and eyesores). Global Illumination is, by definition, an advanced and accurate simulation of light, and it's much less visually forgiving of taking shortcuts than most other lighting effects in real-time graphics.

For these reasons, and again I've mentioned this before, I feel largely unable to make significant improvements past the new cascades update on my own--and let me assure you, either way, you guys will be getting access to this update.

I've discussed the concept of making SEGI freely available and open-source, and you provided very helpful feedback on that idea. I've taken your constructive feedback into account and thought long and hard about it myself. Some of you made excellent points against SEGI being open-source. However, I feel that at this point SEGI being a paid-access beta product makes a promise that, despite my best efforts, I feel unable to fulfill anymore.

I just want to be transparent and honest with you guys. You have helped me and supported me in chasing a dream that I've had since before I even thought I was capable of being a graphics programmer.

So, as sad as I am to say it, I feel that it is best to do something about these problems. There are a few options, and I want to involve you guys in this decision making process.



So, here's the best idea I've come up with. Please, if you have any ideas, I'd love to hear them.

Step 1: Depreciate the SEGI Beta product on the Asset Store. Those who have purchased SEGI will still have access to it, but it will no longer be visible to purchase on the Asset Store.

Step 2: Create a GitHub project for SEGI under the GNU GPLv3 license. This will be the latest version of SEGI (with cascades, blue noise, and all the improvements I've mentioned). Anyone will be able to fork and modify/contribute to the project with the proper attribution. This doesn't necessarily mean that I will stop working on it. I feel that this is the best chance SEGI has of reaching its full potential. I'm not concerned about not making money on SEGI anymore, that was never my motivation for developing it.

Step 3: Anyone who purchased SEGI on August 1st 2017 or later will be granted a refund (which needs to be by request because of how Asset Store refunds work). The last thing I want is for anyone to feel "screwed" by this decision.



I'm sorry that the beta cycle of SEGI didn't go as planned. I really wish it had, and I've been trying to push it in that direction in vain for a while now, but I feel that it is time to steer the project in a good direction rather than wrestle it into a direction it doesn't seem to want to go and have it end up crashing in a worse place as a result.

Please share your thoughts on this. Thank you for your understanding

 

I also think creating a experimental fork or branch where you can implement all of those unpractical features that prioritize quality at the cost of performance (such as directional voxel data) would be a good idea.

From there someone could try and create a fast lightmapping tool.

 

I believe he means August 2017

 

Absolutely! I'd love to do that. Screen-space tracing, in particular, although expensive, brings GI accuracy from voxel-scale accuracy to pixel-scale accuracy. It'll be nice to share those things.

The past 30 days is all I can afford to refund, unfortunately.

 

I am a little upset but understand and am glad you let us know the problem and what you would like to do.

At the moment I can use SEGI for testing and making a scene so much more amazing. I do feel it is usable for a normal optimized project(not VR). Cascading should help a lot.

I would say though please do not let this be dead with a Unity version that breaks it. I understand if you can not add new features but if it is broken in say 2018.1 it would be very good to the people who invested in it to have it working. Understand if it is dead within a couple years as there is a chance, albeit very small chance, that Unity will have something usable by then. It would be good to have SEGI working for reasonable amount of time though as I still think it is great to use.

I am not expecting a lot as it is such a tricky thing to work on for free but it would be amazing if the community could get into it and update it some. So that is a possible glimmer of bounced light

 

Are you sure about GPL v3? Nobody is going to use it in any commercial project ...

 

I think it sounds just right to me. When you are first in to a subject you have to hope the numbers work. You make what you think are reasonable assumptions but surprises can come from anywhere. You can't know until you do it. You took those risks and it must have been really stressful, especially as experiments failed and options ran out. Thanks for bravely tackling this. I think making it a community project is the right thing to do for SEGI. I'm cautiously excited about the future and what we can do with it. Especially with working out how to make it perform in specific use cases. I think that is one advantage of having open source. As a tool developer you feel responsible for making a versatile and simple to use product. Given the limit of current technology we may find tricks to get specific implementations working really well. After all this is how developers have always stayed ahead I think.

Really excited about cascades! Great work overall. Good decision to go open source and the right time to do it.

 

I would like to thank you for the info about SEGI's future, it cannot have been easy to write that and you have my respect.

I am very much into the github/opensource idea. With an eye on practicalities and the future, I would really like you to consider going for a different type of license. For example, in the coming months we are going to see what legs the Scriptable Render Loop stuff in Unity is going to have, and they share that stuff using a MIT license. Whether or not it ever happens, it is possible to imagine people getting SEGI working with this stuff one day, and it would be good to avoid any license compatibility problems.

 

Yeah, as @elias_t said, are you sure about this one? GPLv3 AFAIK requires that you share the source code of the whole project when you use GPLv3 stuff. I'm guessing the intend is that various forks of SEGI remain opensource, but, again AFAIK, GPLv3 forces people to share source code of their games.

 

I didn't read about the license before. Yes I agree this seems a big problem. Making sure forks of the SEGI project are open is fair enough but it should not require an entire commercial project to be open.

 

Maybe this way is moar money-wise:
Step 1: Put latest version of SEGI to Asset Store as usual, e.g. for same 80$.
Step 2: Add full source codes of SEGI to that Asset Store 80$ package (what license is the best for this case - no idea, something that allows forking but not free sharing/ re-selling as GI asset? For example, team from http://developer.space/ maybe very interested to implement this solution and support it by themselves - what license fits best for that? Ask them.). Maybe increase whole price to 100$ because of those source codes - a lot of intellectual work was put in to this.
Step 3: Screw any ideas about refund (why you feel need to do that in the first place? Product works and with latest update it will be even better; this feels like it's maxed out - so, you did everything that was possible, no reason to blame yourself or be ashamed; maybe a little bit for big plans on start but you was honest when you did those - it was not fraud, just lack of experience with limitations of human brain - only puny 15 watts of processing power ... simplifying core stuff as much as possible is the key to build most complex things on that simplified core components ... nevermind XD).
Step 4: Continue minimal support of SEGI (just to make it work on newest versions of Unity). If inspiration strike you (or an Unity 2018.x engine open new ability with huge benefits for SEIG; hardware becomes faster too; your skills improving, etc...) and some other improvements will be done - return to Step 1.

 

This seems good in some ways. Financially it's obviously better and you should be paid for your work. I agree SEGI works and, although it doesn't do everything that was planned in the beginning, it is cutting edge tech that works reliably. I'm just concerned that keeping it under the Asset store package will stop people being properly collaborative as any forks will be commercial property. For example with UMA the UMA Steering Group was formed, and people have improved that project massively. With UMA though Unity funded the original project and also UMA was not breaking new ground in technology so improvements were much easier. I don't know if my concern about collaboration is valid but there it is.

 

I shouldn't say forks will be commercial property. I mean forks will be improvements on something that requires a purchase beforehand. It's like the open source thing in Unity I guess. I don't know how much time people put into open source work for Unity given that its working for them for free.

edit: Actually the more i think about it I think it's not a problem. Also making segi free could prevent other gi solutions from being able to enter the market.

 

Well, Sonic Ether could always release 1.0 with the maximum quality profile considering even 30FPS is considered realtime and none could call him out on it.

After all, this never started as some kind of 'Performant enough for VR' project. Those are some crazy expectations.

Though thinking about it, you could say SonicEther's issue is being too much of a perfectionist. Settling for the middleground with something like CryEngine SVOTI and its many limitations would have been enough instead of making it be a perfect 'all scenarios' GI solution with no flaws which won't happen.

Give yourself some credit and rather than saying 'Out of my skill level' maybe it could be you are trying to do something that is not yet feasible at a decent performance under current hardware.

Even HXGI is dealing with performance by delaying the propagation speed/rate and spreading it across many frames.

Releasing v1.0, making it be compatible with current Unity versions and occasionally popping in if you get any new implementation idea is enough for most as some posters are saying.

 

What? You know what 80$ means in my country? It's a big amount of money in my country and you are saying I paid this amount of money for nothing? That's a big steal! I knew you can't be trusted

I don't expect any new features from you! just make it compatible with new versions of Unity. and don't make it free and leave it. Is that a big wish against 80$?

 

I'm also with Lex4art suggestions. I think you have done an amazing work at a very reasonable price. Please stop further investigations into new techniques and simple don't let SEGI die. Instead please give it support for 1 year from now on with newer versions of Unity and if you find the time maybe add Single-Pass-VR and the ability for 1 time voxelization but multiple cameras.
I think >90% would be happy enough if you'd do that.

 

One solution might be to open a github page but keep the asset store paid. And when you create a new asset you feel like maintaining, you make segi upgradeable to that asset for a discounted rate. Not sure if unity allows though. I don't own segi btw but would love to try.

 

Problem with leaving it on the store is the unhappy messages will continue and I would imaging that is just depressing to get every day. I do like the idea Lex4art posted but I think two things would need to happen so that there is a balance and not a loss for Sonic's potential new regular customers.

-Unless you added features scheichs mentioned like Single-Pass-VR and the ability for 1 time voxelization but multiple cameras than I think lower the price $40-$60(better than free and should support bug fixing). Maybe keep it at $80 for a couple months after the cascade update though as that is a great improvement. Then lower it to a more reasonable price. Unless it featured VR support though. That is golden right now and you should keep it at the current price if VR users were appeased. $80 is a lot of money to some people. Not bad though considering it is a great lighting solution and most people buy new premium Samsung and Apple phones for 10x the cost every year, which are lets say huge pits of money, but have to understand that there always has to be a balanced in pricing and customer happiness.

-Do not promise any new features. It is what it is. Even on the Asset Store page should have a list of its minor shortcomings. If no Single_Pass-VR is added than say no VR support as that will pacify the amount of disappointment with new purchases. If a user comes here and says I need blah blah blah than just say I am sorry but it is a finished product and I am still supporting bug fixes but am no longer adding new features.

Also with this in the future if something clicked and you could actually improve it you can. Should not feel obligated but the option is still there for you.

Again I have to say either open source or stay on the store please do not abandon this project. Understand if you can not add new features but please keep it working with new Unity versions for a little while.

 

This is exactly the reason why devs dont like to promise to much.

Still nice to see the work to you put in this project sonicether. Also thanks for beeing so transparent. Also dont see a problem in devs getting paid for their early alpha so i see no reason in refunding. It was your hard time you put into.

 

Don´t buy stuff in beta and come to cry about your personal issues as if the asset not being ready made your life a piece of ****. SEGI represents a major achievement even at current state. The developer is doing an massive job and this is, in pair with the Other under development GI solution the closest Unity has been from quality, trully realtime GI.

 

As for the future, an addition that would make SEGI a blast is to have a profile where you max the visuals to like an offline renderer, it takes it time to compute ( like some secs) and bakes everything into lightmaps. Then in runtime you could simply run on lightmaped data. Ideally if the lightmaping part is not blackboxed, integrating the lightmaps into unitys realtime/lightmap blending tool would be simply epic. Like Beast 2.0.

 

LOD helps a lot about performance, displaying 10 distant trees of 3000 polys is faster than displaying 10 distant trees of 30 000 polys. For real time games it's a must have when scene are complex and highly detailled.
Perhaps some solution will be available ?

 

Are you own SEGI? then don't decide about our money. I want a refund cause it was just a time loser for me if he going to make this free and leave it on GitHub. he's well known as assets leaver. Which one of his assets is still alive?

Then Why you sell expensive and earn money when you can't keep none of your assets alive? I couldn't believe you going to even leave SEGI! now I'm pretty sure you're doing this on purpose and when you sell good then you leave. cause everyone love the money but nobody likes to work.

I want my money back! it's not your first time it's not even your second time leaving your expensive assets. and I didn't paid 80$ (equal to 15 days of hard work in my country) for a leaved asset you are not even willing to update with newer version of Unity.

 

All the trees in the scene are LODs and contributing to the GI but are not using the terrain system and are just LOD gameobjects. Even the billboards of the last LOD for each tree. The terrain LOD trees in this scene are just small LOD ground cover models that are too numerous to use as LOD gameobjects. They are very dense and would not contribution much useful info to the GI so that is why I hid them from GI and it saved a lot of performance. With a GTX 1070, 1080, or AMD Vega I could easily include them and get 60+ frame rate. Just not preferable with a mid range card.

 

Would be great to get SPSR working

 

Cody, Please check carefully what license to release under

GPL is a legal virus and scare people off.

LGPL is better.

 

There is also, MIT, BSD, or Mozilla Public License.

 

Not my area of expertise, but iirc MIT is the one I see most for cooperative public projects for Unity.

 

Thanks, everyone, for your understanding and feedback! I want to stress again that taking these actions doesn't mean that I'll stop working on it altogether. Updating the asset to ensure compatibility with future versions of Unity will actually be much easier, more convenient, and less time consuming on GitHub than it is on the Asset Store. Also, I'll be able to more easily manage different branches of development/releases. This fact, combined with the potential for new contributors to the project make me feel that the original plan I proposed is the best option and compromise for everyone involved.

Also, thanks for warning me about the GPL license. I'll look into the others that have been suggested.

 

I agree with those who have stated you shouldn't be under-selling yourself. You released an unfinished beta and were transparent about it, and it's people's responsibility to choose whether they want to buy into that or not.

As far as I'm concerned, your product is still the only viable option available right now, and your cascades can make it much more performant. For some people's games it may be perfectly suitable. For others perhaps not. But just updates the info, release a demo or two, describe the performance considerations and limitations, be transparent about how you feel it won't jump forward in any significant way, settle on a price you think makes the most sense, and let people decide of they want to buy it. Your product is still a good product for many situations.

If you want to go open, that's fine too . Don't stress so much, but get me those cascades! Haha.

 

I think going open source is a great and generous idea. I don't want to tell you to give your work away for free, but maybe this just isn't that time for a commercial voxel-based lighting asset. Perhaps the technology needs to advance some more and this could be a great step in that direction. It seems like this is a project that could benefit from lots of people (potentially) looking at it, experimenting, and adding their own ideas to it.

 

Also SEGI really don't have to be perfect :
- Game still struggle with good shadow, and that's two order of magnitude simpler
- Area ligth is still struggling, and that's without any bounce

So a limited ranged full GI is quite an acheivement in real time!

 

Even if you were offering refunds and depreciating the project, I would not want it. SEGI as it stands has been an amazing resource for my project. Being an animator who got into programming, I'm not the best at coding or even understanding all the concepts. But in my opinion option 2 seems hands down the best option. Not only will more people have access, it would also mean that SEGI has a higher chance of improving as it is being worked on by multiple minds instead of one. Also if you go the open source route and still decide to work on it, I'd like to request a donate button so that those of us who want to support what you do with what we have still can contribute. After all, you started this and it is by no means a small feat.

My personal request regarding SEGI's direction would be a more flexible customization in quality. Higher settings, because for my project personally, it is not about frame rate, but a user being able to set up a scene, turn on the GI so it renders, then Disabling it. ( The project is a customizable, realtime studio where artists can study how light affects form, as well as give them the ability to generate references to study from or use to sketch out masterpieces. )

 

as far as getting what i payed for i am a happy customer. (50fps with ultra settings)
i only wish you just keep polishing not adding major new features. and KEEP selling it, remove the beta tag and keep it, SEGI worth the money in this state, not all games are massive landscapes.

i have bought another beta GI and the developer abandoned the project with really unusable product at all.
by far yours are the best solution for unity. don't stop now. make profit you deserve it.
small fixes and tweaks are what needed from now on.

 

I don't own SEGI, so maybe I don't have the full picture on it, but it seems to me like it is working ... it's just not as perfect a solution as the developer was hoping.

So I suggest polishing the UI so that it's easier for users to get the type of performance they're looking for (high performance - medium - high quality), set the update rate, and then take it out of beta and say "Development is done, this is as far as I can take it."

Then keep it on the store and provide updates for each new version of Unity. That seems reasonable.

For those who need a dynamic GI, SEGI is the only option right now.

Of course, if the idea is to move it to GitHub and make it free, that is within the rights of any publisher, whether they are an artist or programmer.

 

OR option 4:

Unity hires him or buys SEGI off him so that we can have realtime GI built in, as well as the ability to set up lighting in the editor and then hit 'bake' through SEGI for higher quality light maps, baked in realtime! I'm sure it's more complicated than it sounds, but the point is SE deserves to have Unity recognize him for what he's developed. ( in my opinion )

 

I guess this is a community-specific phenomenon where it is collectively considered OK for a developer to release a product, charge $80 for it, promise numerous features, basically avoid supporting it for over a year, then deprecate it after collecting payment. The fact that this is almost exactly what happened with SESSAO, really only makes it worse.

I'm not sure why others are not of this opinion, beyond the fact that Cody seems to be a good guy (?) and of course, real-time GI is a hard problem and it's an achievement in and of itself to have approached a solution at all.

I suppose, as a VR developer, I am less forgiving because SEGI really is not satisfactory for that particular use case, but I am very disappointed in the way this asset has panned out, especially for the price. I wouldn't say it's at Livenda levels, but I'm definitely not happy about it, or SESSAO for that matter.

 

I guess these are single developers and not bankrolled. So I always approach these things as curios for fun, and sometimes useful, but not guaranteed. For guarantees, I head to first party: whatever Unity's done, or whatever I've done.

Unless it's @Eric5h5 who for some reason just supports his products for at least a few thousand years.

 

I have to say, I don't know any of the people on this thread personally, but I see their contributions on many different forums, and as a result, I respect many of them. That being said, I cannot deny you, @hippocoder, you have been such a dedicated personality on the Unity forums, so I have to assume you have a more mature perspective on these issues.

 

I feel a stab of anger for every asset I wasted money on, I'm human. It's just what else could I do? run around shouting like a little boy?

 

Reply
VR requires minimum 90 fps... I think that expectation was unrealistic to put it mildly. When I first started seeing talk about VR on this thread I was pretty surprised.
Did Sonic say it would work well with VR?

 

Well, people buy them, and they tend not to need many changes with new Unity versions, so it seems reasonable.

--Eric

 

My take: put the cascades, add full source code, raise the price to 100, and keep it alive. Plus: add a few easy to tweak features in a UI, like the ability to set the amount of refreshes per frame, little things like that, and call it a day. SEGI is the only option at the moment, and a success - it's kind of incredible that a single guy beat the whole team of Unity and their crappy lighting system.

But do push that cascades update, our game is counting on you ;-)

 

I don't want this to die. I guess it has if you(sonic) have decided to just go through with the open sourcing, and truly believe that further improvements are beyond your capabilities but I knew I was buying into a beta, and it works.
I would rather use this and its limitations than roll with enlighten(which takes forever to get half decent results with). In my mind I'd like to see a private source access given to customers who might want to fork their own modifications, you could roll up whatever you wanted out of this and just continue development even if its at a glacial pace. I personally would rather you fix things for image quality and screw speed, hardware will always be improving(I envision the next round of VR devices driving a renewed need for speed on the hardware side of things).
Anyway that's just my opinion.

 

I don't think this argument hold here, the community coerce the release despite the author warning, we were ready to take the hit because the demo where simply amazing, as is. Our fault not his, the alternative was jumping to this story conclusion but with nothing in our hand, not even github, I would say it's a net win overall! Now we have something to study and expend, maybe even adapt, from generic catch all solution to project specific optimization and hack! Which might serve VR better in fact...

But then there is also hxgi, de can only win...

 

How about this. Opensource it but commercial license to only those with proof of purchase. So immediately to those with a unity invoice of purchase.

Take it off the unity store and sell the license at a personal website. You can even push this money to charities.

This opensource commerical license arrangement is something like what emguCV is doing.

 

I think the whole point - why Unity Tech choosed Beast/Enlighten - is broader hardware compatibility. Imagine they would offer a realtime GI solution like SEGI, HXGI or even NVIDIA's VXGI - complaints incoming. Beast/Enlighten just works on (almost) every hardware - no matter what computing power the end user has.