Not sure if this is actually the case, but in your last image it seems like the GI isn’t respecting the normal map on the tree. Or perhaps it is but it’s not granular enough to notice?
There’s actually a bunch of issues with that.
It uses the bent normals derived from the AO which is what’s causing what ur seeing. I still need work on denoising / and filtering for the bent normals specifically. As off now I’m just using bilateral blur and some spatial filtering and basic temporal reprojection as a place holder. So the normals on surfaces get smoothed to crap atm. I should also be taking the AO factor on screen and adjusting the lighting as well. I also need to account for unitys ambient light factor better. It was a really quick toss together, lol. Given I set up all the data generation and wrote the effect up in 4 hours I thought it looked alright, lmaoo.
Good eye though. 
I’m not going to put to much attention on the bounce lighting at the moment, but I figure once I get nice filtering for the AO in place relatively soon, those issues u pointed out won’t be apparent. Could actually eliminate then all together by just using the normals in the Gbuffer already. But ya, indirect lighting is def a work in progress still, haha. '^^
3 Likes
Is it possible to bake SDF data on objects made of multiple modular pieces such as this one?
As you can see, the building is constructed from flat planes.
Ok cool, was just curious. Good to know it’s not a fundamental limitation. I’ve been looking at realtime GI solutions for like, 6 years(?) now and whenever something new presents itself I’m immediately interested lol.
Yah, it is! At the moment u just whip any game object into the source gameobejct field in the converter, then hit the generate distance field button after tweaking the scale and offsets if needed. It does a pretty good job of figuring out most meshes inside really complicated game-object hierarchies.
So fast update, im basically settled now and will be around a lot more to anwser questions and what not. I finished most of the clean up this weekend and the stuff thats holding me back from getting this out to testers. So hoping in a week-ish I can start getting this out to people and give this thing one last round of polish before submitting for publishing.
So I ended up opening pandoras box so to speak. I was curious what unreal was doing exactly so I looked a lil deeper at the source. It turns out the bent normals are for adjusting skylight stuff. I was under the impression they were more akin to unigines screen space gi’s deal. So sorry for the naivety, i think i can support something still more like i would be interested in though but ill save that.
But ya, It looks like unreal actually down samples at 1/4 res then upsamples to full which is mind boggling… Anyways I think i ended up with something a lil better then UE4 unintentionally. (Posted UE4 DFAO shot for comparison at the bottom)
So i put in a new set off filter built from the ground up. This was always my weak point so sort of strapped up over the weekend and experimented quite a bit. I ended up with 90% of SVGF. The re-projecting and variance calcs are done the wavelet filter works for the most part, just having a bit of a time actually using the variance to adjust the blur. For this i actually thing SVGF is too much for, but nvidea’s reprojection bit is pretty awesome so ill probably set something up that’s less GPU intensive for denoising this stuff specifically with that, but it should open doors to redux ray counts introduce interleaved rendering for more ray count reducing. I unno. For w.e reason i have a feeling this is more the direction UE5 is going with SDF and i figure it would probably be worth my efforts to follow suit. Also runs surprisingly well… the blur costs more then the AO tracing… lol.
Unity DFAO
So just for the sake of comparison; Here’s a good look at the AO now; All ray traced in unity. I def need to work on my edge avoid metric to get the trees to blur a lil less but it’s alooooot better then anything ive posted prior.
Unreal Engine DFAO
And here’s UE4; I tried to find a terrain / forest scene thing. And yaaah. Anyways, think im solid enough to release this. Was having anxiety about my map magic terrain not looking like the UE4 kite engine demo and i guess i went to far with this in the end… oh well haha.
Not sure how relevant this is but, i think ill probably focus on hardware ray tracing and indirect lighting / reflections via SDF shortly after publishing this. If SVGF works out well enough, could really do a lot easily. I also think i can adjust the AO calcs to works to alot like SEGIs to sort of handle the indirect and AO stuff cumulatively. I think i can pack the colors as well so the cost to sample SDF data would be basically the same… Anyways future looks bright!! 
Probably gonna redo the entire main post again and really finalize that throughout the week; If anyone is interested in testing this stuff, send me a DM, im looking for around 10-20 people working in a variety of scenes. Specifically standard rendering for the moment. HDR can be supported but I’m holding off before opening that can of worms until I have something out there before this snow balls into “just one more feature!” oblivion. I know HDR is pretty important, but this is gonna take some TLC so i figure I can probably have HDR and support for bounce lighting and reflections a month or two after the initial release and that’s gonna be the focus unless something more demanding comes along.
Anywhooo, stuff looking alot better. Codes alot cleaner… I tied most of the lil loose ends and should be able to manage the rest of em through out this week. Just wanted to say basically wrapping up, haha. And i think all the ground unreal covered aside from messing with ambient calcs and stuff has been covered; Clip map, Instance Cone Tracing, and Height field Derived AO and Instance Cone Traced DF Soft Shadows.
11 Likes
hello sir
looks really nice !!!
will you make public properties we can access for custom shaders ?
Wow, this is looking nice. I was playing with implementing a SDF AO solution, but I would rather use an asset that’s already done if that’s an option. We are targeting VR platforms using a URP spinoff. Any idea if you plan to support a forward based pipeline or at least have a custom buffer mode?
Any updates please? This looks really interesting and it would be a shame to see it abandoned!
Did this really end like anything else, is it really next dead project?
It’s only been a few weeks since the last update - have some patience, developing assets takes time.
3 Likes
really waiting for it
its realy dead, isnt it?
Any updates on this ?
Looking forward to this one. Good things take time!
Just a heads-up… I am NOT saying with this dev, I have been around
Unity since 5. One pattern I have found is a DEV that does this and
does not log in for months at a time.
IF* the asset is ever released, chances are it will either not be kept up,
or deprecated in a year or 2.
Just saying careful on this one.
Hey Everyone,
I just wanted to give an update and apologize for going AFK forever.
The response was also a lot more overwhelming then I had initially expected and I tend to Sean Murray / no man’s sky under any public pressure admittedly so I apologize for that as well. In other words my ambition can get the best of me at times.
Kennth’s observation is very valid.
It’s easy for things like this to slip off to the side and dissolve into vapor-ware dealing with the day to day realities of work, personal life, etc. The road to hell is paved with good intentions and the logistics to manage something this big can be insane especially if you occupy a full time senior position for a third party, lol…
I had ended a 4 year contract to work on a pretty time demanding project in November of 2021 and just needed a break from computers in general after it for a number of reasons.
As of now and throughout most of the winter I’ve been organizing my own start up and have been actively updating the code base for the past few weeks integrating and modifying this system in it, as well as migrating the lighting system itself to work with SRP stuffz (finally) and thought it might be worth giving an update on the project as whole.
As of now I’m hoping to get it stable, tie off loose ends and get it to a point where it wouldn’t be so daunting to realistically attempt to release this either commercially or for free on github or w.e. In either case I do intend to release the code in some capacity, what direction ill be going I’m still feeling out at the moment.
So in vein of anyone expectations I will say I’ll be taking this slow and more humbly from here. It’s not so bad to manage something that’s complete but in it’s current state it is hard to estimate as to when something will be done or make any promises in general in regards to releasing it.
Anyways, main reason for posting this aside from giving an update on the current state of this; Tomorrow, ill be setting up a discord server for managing a number of personal projects. I’m planning to make the code base public from with-in a testing channel on there. Anyone can join it. PM me and we can discuss testing channel access to gain access to the files and what now if you’re feeling brave/ambitious enough to mess around with it in it’s current state.
Ill update the main thread post with the discord server details when it’s all set up tomorrow.
Cheers and thank you to everyone who did support the idea of unreal-ish lighting procedures in unity 3D.
Regards,
SuperDrZee
27 Likes
1 Like
Hi everyone!
I created a DFAO system for my gi lighting system, if you are interested here is the link to my github:
My main prio is still Nano Tech, but I will continue to work on the other tools too, if the crowdfunding runs well:
11 Likes
Big Update
So ya I opened the code base back in march and was reminded of why i was hesitantly to confidently say, ‘This is something I can realistically manage’. pretty fast, lol.
In that regard though, I decided to just keep my mouth shit and I ended up spending most of the spring fixing up/cleaning up the system and I am personally happy to say I will be pursuing developing this at a full time capacity and I have just finished the first stable version that can be tested and I’m looking to get things going again.
I’ve started a server to streamline testing, inquiries and other stuff related to the project.
Over the next week i’ll be assembling a small-ish test group of around 20 or so users. If you would like to be a part of this, join the server and PM me, the owner directly with details about your hardware an your target rendering API / unity rendering set-up.
Through out the evening I’ll be editing the main article to reflect all the changes as well. The quality of the effects is about 1000 times better. (No exaggeration, those nasty precession errors.) The cost to run these effects has been optimized like crazy. I’m basically at where unreal is now in terms of rendering costs. Big things anyways.
Lastly, I would like to apologize again. To be frank; OH boy was this a messy start. I will say now I’m looking at this like a blank slate, the code actually works with no rubber bands er band aids, my life isn’t imploding and I can say to myself that I have the time to manage and support this.
With that said, I will be having my partner help me with managing the unity forum stuff, so nothing gets missed. I will also be taking sometime this evening to finally PM each every one of you, whos contacted me in the past. It was a little overwhelming having to respond to the same msgs 100 times, saying ‘just give me a couple months’, but yah, ill be making sure everyone gets a link to the discord and a proper opportunity to discuss testing in a more streamlined fashion.
I will say as well, if you want to get ahold of me, discord would be the place to do it, if your looking for an immediate response. I’m on there 12+ hours days, even on weekends and would be my preferred method of contact.
Anyways, system as far as the baseline for a release goes is done!! Sorry again for being an asshole, I honestly didn’t think last winter it would ever make it here though, and i really want to make sure my shit was completely together before promising anything ever again.
So let this post serve ultimately as my way of saying, i’m taking on supporting this and potentially other products as a full time gig vs trying to boot strap it with my head cut off. Hopefully with that, and over a little time I can offer some assurances in terms of my abilities to support and manage this.
Anyways, can’t wait to show off where it’s at now. Everything looks sooo much better.
11 Likes
How does it compare to native unity shadows in terms of performance and visual quality? I would love to see some stats and pictures. I would be interested in testing this, but I am locked into the HDRP project currently. Soft, performant shadows is what our project needs the most. Cheers!