I’m not really understanding how LightImportanceVolumes work.
I’m working on a part of a scene which is lit only by indirect lighting (the scene is basically an apartment room). Indirect lighting intensity of the source light is quite pushed here, which helps in seeing the problem I’m talking about, but the same issue is visible with standard intensity values.
The room (and only the room) is completely contained in a single Lightmass Importance Volume.
Building the lighting produces dramatically different results (both in tone and luminosity, and in the relative luminosity of different parts of the walls) moving the Lightmass Importance Volume of just few inches left or right (leaving exactly the same objects inside):
Does this kind of problem happen to meshes partially inside and partially outside in your experience, or it may affect also meshes completely inside if there are meshes crossing the boundaries?
In any case, in my test above all the meshes were inside the volume.
Apparently, if the volume has one edge shorter than the other (e.g. is not cubic), movements along the shorter edge have more impact.
It is kind of a problem, randomly sometimes I get a lot of artifacts and sometimes completely clean surfaces, depending only on the position of the volume:
The problem is, I’d like to avoid to iteratively try to find the best setup (it is very time consuming, since I have to rebuild the lights everytime, and worse of all, if I find an acceptable solution and then experiment different setups I have no way to come back to the acceptable solution, unless I save the map as a different level)
Please notice that the artifacts in my case may more visible due to the high indirect light contribution, but the same problem arises in standard setups.
…so, I would be interested in the following questions:
a) why the lighting is changing moving the volume of even few inches? even if the size of the volume doesen’t change? how does it work?
b) what is the optimal setup? a volume as small as possible? should it be cubic? should it be centered on my room? symmetry is important?
c) is there a way to note down position and size of the volume without saving the whole level? (should I copy and paste it to a text edit and back in the project?)
d) Directional lights need to be placed inside the volume?
and bonus question:
e) what are those splotchiness due to? They appear in low lit area, and I can fairly exclude e.g. in this specific case that the cause is linked to UV layout or lightmap density
The same issue is still present in 4.5 preview.
Since there have been no hints about why this is happening, I guess it should be considered a bug. Would it be possible to move this thread in the bug reporting section please?
Would it be possible to move this thread in the bug reporting section please?
Done!
Sorry i couldnt be more help on this. There shouldnt be such drastic differences between light builds as long as meshes are completely inside the volume, so i dont know what is up with it.
Ask around in the rendering section of the forum if you cant get further help here.
Now that you mention cache…My experience with messy lightmaps for partially included meshes goes back to UDK, and not related to this problem but some other lightmass artifact(which i cant even remember anymore), i used to clean the cache(open up swarm > top menu bar > Cache > Clean Cache) in Swarm before building the lights. See if it helps.
And i dont really know the technical reason behind it but my guess is(or in my imagination ); since the photons are distributed inside the lightmass importance volume and part of the mesh is outside of this volume some error occurs while preparing the lightmaps and that causes the blotchiness. But of course, this is not the case in your situation apparently since everything is inside the volume, so see if cleaning the cache fixes it!
… by the way, what is the technical reason why meshes only partially inside the volume may cause this? Maybe understanding a bit better how this works would help me eliminate the problem.
In my experience, lightmass calculation is pretty buggy: sometimes I get plainly wrong results (like the ones in this thread, or strange impossible colored haloes in other areas) and then restarting the editor and rebuilding the lights leads to a more normal situation. Is there a cache or something? can I force a full rebuild? What are the tools I can use to investigate these problems?
Hmm, i’m using Windows and i havent touched 4.5 preview yet. Cant you just build the lights so that Swarm starts itself? On Windows Swarm starts and minimizes itself when you start building the lights, so you can cancel the build afterwards > clean the cache > rebuild. Or search for SwarmAgent in Engine folder and start it manually to clean the cache.
Umh… how do I launch swarm in OSX? I can’t fine the launcher in Engine\Binaries
Any other way to clear the cache? in 4.4 there was a Saved/Swarm/SwarmCache folder, but I don’t see it in 4.5
"… by the way, what is the technical reason why meshes only partially inside the volume may cause this? "
From Lightmass documentation:
“Lightmass emits photons based on the size of the level, so those background meshes will greatly increase the number of photons that need to be emitted, and lighting build times will increase. The Lightmass Importance Volume controls the area that Lightmass emits photons in, allowing you to concentrate it only on the area that needs detailed indirect lighting. Areas outside the importance volume get only one bounce of indirect lighting at a lower quality.”
-So the Lightmass Importance Volume designates the area where full photon emissions and their subsequent bouncing will occur (about off of 5 surfaces). Areas outside of the Lightmass Importance Volume show photons bouncing one time off one surface. So the purpose of the Lightmass Importance Volume is to only have to spend time rending complex lighting where “important.” Otherwise it would try to render with high definition lighting in every direction for infinity.
Hence, meshes that exist both inside and outside of the Lightmass Importance Volume “confuse” the editor which can cause unexpected and undesirable results. -If your scene is small enough, I suggest encompassing the entire scene (all meshes) within the Lightmass Importance Volume while keeping the volume just large enough to acheive this.
Thanks for the info! In my case, anyway, all the interested meshes are completely inside the volume. I do believe this behaviour is related to some bug in lightmass…
Sorry, I did over-look where Jacky and you discussed your meshes were completely inside the volume -I just was hoping to “shed some light” on the lighting volume itself. I am looking into this as a bug.
Never mind, any help or hint is always welcome I’m trying to further investigate the issue… in the screenshots above it was amplified by using strong indirect lighting intensity, but it is definitely visible also setting this to 1. Plus, it may be linked to nearby translucent objects… making them dynamic instead of static seem to affect the result also. How translucent object (eg. plane of glass) are supposed to influence lightmass?
Also, when building a room, is it to ok to use 2d planes instead of boxes for the walls? The room in my case has an outer shell to prevent light leakages, and in any case replacing this specific part of the room with a boxy mesh didn’t change the issue, but may this be linked to the issue? Thanks!!
I would recommend trying two to five perpendicular planes (like a hollow box with missing planes you want to be open) that belong to one mesh rather than boxes. Be sure you have a separate UV map for your light-map and set this in UE4 accordingly.
2.You are correct in assuming this may be linked to nearby translucent objects so I would try step one without them present first and if that is successful, add the translucent objects back into the scene.
Thank you, 1) is more or less what I’ve been doing so far
as concern 2) removing the translucent objects (making them invisible and dynamic) changes the situation, but does not solve the issues.
The room I’m working on is lit just by indirect lighting, and the result seems to be pretty unstable. From my latest tests, similar issues arise not just moving the LightmassImportanceVolume, but also moving very slightly lights or strongly lit objects.
Even when the result are not so dramatic, very slight movements of these objects can result e.g. in the ceiling getting much darker than what it should (I mean, such small movements do not justify so big changes)
Also, the tinting problems seen above may be linked to some caching of lightmass, restarting UE4 Editor in some cases improved the situation
I’ve tried to create this issue on our Mac here but to no avail -but likely I’m not recreating your situation accurately. So if you feel this is a bug that needs to be addressed, please provide more information about your Mac’s specs.
You can find this in the Apple menu >About This Mac>More Info> Add the information from this window (serial # not necessary).
We also need the following information from the “System Report:” Apple menu >About This Mac>More Info> System Report>Hardware (All information listed in the “Hardware Overview.”) You can cut and paste this information or provide screen shots, whichever is more convenient.
This information is required in order to file a bug report. While I cannot guarantee this will be addressed, I have recently seen similar bugs fixed so your input could result in reconciling this issue.
); since the photons are distributed inside the lightmass importance volume and part of the mesh is outside of this volume some error occurs while preparing the lightmaps and that causes the blotchiness. But of course, this is not the case in your situation apparently since everything is inside the volume, so see if cleaning the cache fixes it!
I’m trying to further investigate the issue… in the screenshots above it was amplified by using strong indirect lighting intensity, but it is definitely visible also setting this to 1. Plus, it may be linked to nearby translucent objects… making them dynamic instead of static seem to affect the result also. How translucent object (eg. plane of glass) are supposed to influence lightmass?