@ - is able to remove all the bugs with good quality GI, also it’s clearly evident that he is not willing to share any of his workflow (Totally understandable). I am definitely going to read the document he has posted but wondering, do we all have to invest that much of time in understanding something that is fundamentally basic in producing any kind of work from Unreal Engine. I am totally willing to give all my time, but where are the documents, tutorials, scene break downs? I can read a book about photon mapping in Unreal Engine - But Need a Book for that.
I humbly request you to explain the working of Lightmass as detailed as possible with visual examples. Video Tutorials would be heaven sent.
I am a Modo user, that community is very cool everyone shares openly everything. The funny thing is people find out eventually and to think that someone may take work from you if they find out your techniques is nonsense. The chances of someone competing directly for the same client with someone else using UE4 are slim. If that’s what the secrecy is about?
No, of course not. The paper linked is for rendering engineers like myself who implement GI solvers. I will test out the scene monday-ish with the original settings and Rafa’s settings and see what I can find.
Well, sells lots of scenes on his website, so there is no secret about that…also he made a tutorial on UE4 Architectural scene setup on Evermotion and also gives some advices around…also che Realistic rendering scene which comes with UE4 is a very good starting point…and the scene made by koola available on the marketplace…
There is no “Universal Lightmass solution” for every scene, otherwise there will be one parameter to set.
While working on the competition ( ), I experienced certain unpredictability of result depending on position of Lightmass Volume. The position seems obvious, it should encompass the scene. But how much ? By atleast 1 or 10perc. ? Seemingly miniscule movements of the volume, seemed to often drastically alter the splotchiness. It was so random I later didn’t touch it even if needed. I got the best result with the very smalled encompass I could get.
I also sort of don’t believe universal settings for perfection can be found with photons as secondary GI solver. MentalRay tried for years and it is what it is when it comes to complicated GI (of architectural spaces), requiring very unique treatment depending on scene. Which is also why other renderers brought more advanced secondary solvers (for example Vray’s LightCache, while also not perfect esp. when it comes to lightleaks, it’s far more evolved than photon mapping, and also has mode to work in world space and is often used for baking down scenes).
[DevOptions.StaticLightingProductionQuality]
NumShadowRaysScale=32
NumPenumbraShadowRaysScale=64
ApproximateHighResTexelsPerMaxTransitionDistanceScale=9
MinDistanceFieldUpsampleFactor=7
NumHemisphereSamplesScale=64
NumImportanceSearchPhotonsScale=6
NumDirectPhotonsScale=32
; Decrease direct photon search distance so that we will have more accurate shadow transitions. This requires a higher density of direct photons.
DirectPhotonSearchDistanceScale=.5
NumIndirectPhotonPathsScale=32
; Need a lot of indirect photons since we have increased the number of first bounce photons to use for final gathering with NumImportanceSearchPhotonsScale
NumIndirectPhotonsScale=64
NumIndirectIrradiancePhotonsScale=32
; Decreasing the record radius results in more records, which increases quality
RecordRadiusScaleScale=.45
InterpolationMaxAngleScale=.75
IrradianceCacheSmoothFactor=.75
NumAdaptiveRefinementLevels=3
AdaptiveBrightnessThresholdScale=.25
AdaptiveFirstBouncePhotonConeAngleScale=2.5
Render time: 4 hours
On machine with i7 950 3.07 GHz RAM 24 Gb
I’m afraid to imagine what will happen to the big apartment!!!
I do not understand these parameters. Just do not have time to study the photon mapping. He began to study and my head was going to explode soon. It is a science. Examples would be the settings for daytime sunlight, cloudy, evening. Why not make the presets. How well do creative work, not to study the photon map. I really liked the graphics in Unreal Engin 4, but once started architectural visualization, faced with the mysterious file BaseLightmass.ini
I can’t believe that koola is the only one in the entire comunity which spent 1-2 days messing around with the baselightmass.ini settings…and as seen even if he didn’t understand everything the results speak for themself.
Regarding each single parameter on the .ini file, well if DanieW is willing to share some knowledge would be great.
In my tests, this is the only parameter you need to tweak in baselightmass.ini to remove the skylight “splotches”.
And stop using my old BaseLightmass.ini, like I already say, it’s roughtly just random values^^.
You don’t need to boost all the values if you just light your scene with a skylight.
With the spot/reflector technique (that simulate a skylight with more than 2 bounces) you need more tweaks (and more render time) to have a good quality.
@Koola - What a perfect timing, Now we got all the fuel we required to boost Lightmass learning. If possible kindly have a look on the level with spot lights for the light leak errors.
I have to confess that I am already using NumHemisphereSamplesScale=64 Value on all of my latest projects without touching other values much (I found this during my experiments that this value is greatly improving the results). But again all I know was this value is improving my results but how? and why? are the questions which still need answers. Light leaks are still unresolved on .1 Value in Static Lighting Level Scale.
@ - I am becoming desperate to know your findings. My hopes are super high. Kindly provide explanation behind all this.
Im wondering about this as well, especially cranking up the Indirect Lighting Quality adds sooo much render time.
And I actually found that a value of 2-3 gives me a much cleaner result than anything over that.
Generally you want all surfaces that need good quality to be contained, plus 10m more or less. The different results from changing it slightly are probably because it redistributed the photons, and there was enough noise from photons in the first place. In other words, there was significant noise present and changing any parameter slightly will shift the noise. The volume is just one of those parameters.
I did a quick search, their LightCache is what’s known as Importance tracing. It’s not a replacement for photon mapping, in fact you can combine the two nicely. Importance tracing = trace from the eye, photon mapping = trace from the light. Many lighting situations, like bright sun hitting a small spot and illuminating the whole room, require a mix of both. In the case of Lightmass though we are building lightmaps, there is no eye position, so it’s not applicable.
I’ve done scenes with .9, 100, 10, .66 and it took round 7 hours on 2 machines. Not too bad! Under 10 hours is ok-ish but over 10 hours it’s a bit long. Koola might use a bigger renderfarm who knows!
