[Mac] Shadows/Snapshot Issue

[Blarn Boo]

Hey!

First time posting here, hoping someone could either tell me what I'm doing wrong or confirm whether or not this is a bug.

Whenever I take a snapshot, the resulting image is never identical to what the second life viewer is showing; particularly with shadows, ambient occlusion and DoF, in short the effects from DoF and ambient occlusion that I'm seeing in the viewer are not present in the snapshot image or are greatly reduced. I have to really over apply DoF to see an effect on my snapshot, and no matter how I fiddle with ambient occlusion, its effects on shadows isn't evident on the snapshot image. Shadows will look smooth and accurate on my viewer, but a snapshot will produce random shadow artifacts and jagged edges as though ambient occlusion wasn't turned on. 

I've tested this on Alchemy's project viewer, LL's viewer and Firestorm (example below is from Firestorm. I would try Black Dragon as well but have yet to set up Parallels for it.). Every time I have graphics on Ultra, save for draw distance which I minimize. Quality of shadows is maxed when I can depending on the viewer, specifically on Firestorm I've tried to max/fiddle with the extra shadow and ambient occlusion settings to no avail. I will add that I'm on a M1 iMac and am getting decent performance regardless of settings, so I don't believe it's because of hardware but I could be wrong.

 

First image is a Gyazo screen capture to attempt to show what the viewer is displaying, smooth shadows and no weird artifacts. Second image is a copy of my high-res snapshot, shadows are suddenly not smoothed out and appear in random positions, as though ambient occlusion was not turned on. 

 

Edited August 2, 2021 by CoralineO

[NiranV Dean]

All screen effects requiring additional finetuning for higher resolutions is not a bug, this is normal behavior.

All your settings are made for the resolution you see on screen, which is the original window resolution, taking a snapshot at any other resolution requires finetuning these settings for that new resolution. In other words, you have to configure your settings for the target resolution. That's why there is a preview window.

All Viewers show this behavior and this is totally normal due to the way these effects are calculated (absolute rather than percentage based).

Note that higher resolutions require much higher/stronger settings which in turn need higher hardware requirements which multiply with the already increased hardware requirements of higher resolutions, this can and will quickly lead to a crash if your GPU isn't at the very least a GTX 1060 (6GB) and up. I tested up to 12.000 x 6.000 shots and even those will eventually lead to a crash if done often enough.

[Rolig Loon]

If you ever think you have found a bug in the viewer or the SL server code, the place to report it is in the JIRA >> https://jira.secondlife.com/secure/Dashboard.jspa  . Follow the advice on the landing page and at http://wiki.secondlife.com/wiki/Bug_Tracker .  Before you submit a report, do a careful search to see whether someone has already filed a similar or related report.  If so, it is wiser to add your observations to that report instead of opening a new one.

JIRA reports are reviewed and prioritized for attention by technical staff. You should receive acknowledgement when you post a submission, but will probably not hear anything further about progress until the issue is resolved (which can take some time -- maybe years, depending on the priority).  You can always log back in to the JIRA to see the report's current status or to add more information.

In any case, this Answers forum is not the correct place.  Lindens rarely come here.  When they do, they are not looking for items that should be in the JIRA, in Support Cases, or in Abuse Reports.  This is a resident-to-resident forum.

ETA:   The JIRA is for reporting issues you have encountered in the SL viewer or server, not things observed in third party viewers.  If you suspect there is a problem in a TPV, submit a bug report to their support team.

Edited August 1, 2021 by Rolig Loon

[Blarn Boo]

Oh great! Maybe I'll find a solution somewhere there, if not maybe another user here will tell me the probably painfully obvious mistake I'm making somewhere. 

Thanks!

[Blarn Boo]

Just continuing this in case someone else has this problem and stumbles here.

I'm still not sure what causes the issue, but I have found something that helps reduce the shadow artifacts from appearing. Photo on the left is with unconstrained proportions, specifically high res of 7860x7860 (like my original example), photo on the right is with constrained proportions 7860x4320. Less shadow artifacts if any, and the DoF effect is more evident, now idea why and with complicated shadows such as my previous hair example it still gets messed up. Seems the higher I push the resolution of the snapshot, the more DoF/Ambient Occlusion/Shadows struggles and differs from what the viewer presents.

[Whirly Fizzle]

This is a known "bug": BUG-216400 - DOF settings ignored on hi-rez snapshots

This bug is closed as a duplicate but the original bug report isn't public. LL have imported the bug though.

[NiranV Dean]

8 minutes ago, Whirly Fizzle said:

This is a known "bug": BUG-216400 - DOF settings ignored on hi-rez snapshots

This bug is closed as a duplicate but the original bug report isn't public. LL have imported the bug though.

A) This is not a bug. This is intended behavior.

B) The description of that bugreport is so off and the pictures even show that whats being said is not the case. You can very clearly see Depth of Field working on both shots. Infact Depth of Field is exactly the same in strength on both shots, the only difference is that the other image is much bigger thus the effect is diminished relatively seen to the total size of the image.

C) It's sad how little people actually look at their pictures after taking, they would have seen that Depth of Field is there.

[Madelaine McMasters]

16 hours ago, NiranV Dean said:

A) This is not a bug. This is intended behavior.

B) The description of that bugreport is so off and the pictures even show that whats being said is not the case. You can very clearly see Depth of Field working on both shots. Infact Depth of Field is exactly the same in strength on both shots, the only difference is that the other image is much bigger thus the effect is diminished relatively seen to the total size of the image.

C) It's sad how little people actually look at their pictures after taking, they would have seen that Depth of Field is there.

A) No, it really is a bug, unless one intends to misunderstand RL photography.

B) Think about real cameras, which is what DOF attempts (incorrectly) to model. Changing image resolution/size is analogous to swapping camera bodies out from behind a lens. The resolution of the body's imager has no effect on DOF. Unfortunately, because SL DOF is computed incorrectly, you can't get very low effective F ratios (shallow DOF) at high resolution. This is because the DOF settings seem to control the blurring radius in pixels, not degrees. You can see this quite easily when resizing the viewer window. If DOF worked properly, shrinking the viewer window would do nothing to the view but lower the resolution. Instead, the DOF decreases along with the window size. As you increase window/image size, DOF increases. In the extreme, an infinitely large image would have everything in perfect focus, regardless of DOF settings, because the finite pixel based blurring radius would be an infinitesimal fraction of the image size. The two images shown in the bug report should be identical, except for resolution. They are not.

C) It's amusing how little people actually try to understand what they're criticizing before they criticize it. Thank you for the irony!

Edited August 7, 2021 by Madelaine McMasters

[NiranV Dean]

15 hours ago, Madelaine McMasters said:

A) No, it really is a bug, unless one intends to misunderstand RL photography.

B) Think about real cameras, which is what DOF attempts (incorrectly) to model. Changing image resolution/size is analogous to swapping camera bodies out from behind a lens. The resolution of the body's imager has no effect on DOF. Unfortunately, because SL DOF is computed incorrectly, you can't get very low effective F ratios (shallow DOF) at high resolution. This is because the DOF settings seem to control the blurring radius in pixels, not degrees. You can see this quite easily when resizing the viewer window. If DOF worked properly, shrinking the viewer window would do nothing to the view but lower the resolution. Instead, the DOF decreases along with the window size. As you increase window/image size, DOF increases. In the extreme, an infinitely large image would have everything in perfect focus, regardless of DOF settings, because the finite pixel based blurring radius would be an infinitesimal fraction of the image size. The two images shown in the bug report should be identical, except for resolution. They are not.

C) It's amusing how little people actually try to understand what they're criticizing before they criticize it. Thank you for the irony!

A) This is not RL photography this is 3D rendering and its not a bug. A bug would be unintended behavior which this clearly is not. Unintended would be if they designed it to scale with resolution but it doesn't. The shaders were NEVER designed to scale with resolution, their inputs (settings) are absolute = intended.

B) You expect them to look the same because you don't know how the underlying rendering works, you expect it to scale with resolution, which its not made to do and was never made to do. Making them scale with resolution and always look the same will be nigh impossible unless you artificially upscale them you'll always end up with a slightly different image due to having more pixels which result in finer details which in turn change how the final image will look like. Post Processing glow is one very big issue here as Glow depends on every single pixel and its brightness according to your settings, having more pixels can mean more pixel are together which will have more glow in total or in worst case details, like glitter or sparkly reflections (sun on metal for instance) will be finer and thus occupy less pixel area compared to the smaller image, this results in glow potentially "vanishing" because there are not enough pixels bright enough and close together to fire off glow.

C) I do understand both RL and SL photography, in case you missed it i have been doing photography since 2007 (the very little that was.... "fancy") and i've been making a Viewer since 2011 that is made for photography. I'm sure i should be well prepared enough to criticize this. Besides, i'm not really criticizing anything, i'm saying that the bug report is basically a false report and should instead be a feature request to allow effects to scale with size which would most likely improve the chances of this ever happening. I don't see this ever being "fixed" as a bug. They will most likely just look into it and say the same i did.

[Madelaine McMasters]

6 minutes ago, NiranV Dean said:

A) This is not RL photography this is 3D rendering and its not a bug. A bug would be unintended behavior which this clearly is not. Unintended would be if they designed it to scale with resolution but it doesn't. The shaders were NEVER designed to scale with resolution, their inputs (settings) are absolute = intended.

B) You expect them to look the same because you don't know how the underlying rendering works, you expect it to scale with resolution, which its not made to do and was never made to do. Making them scale with resolution and always look the same will be nigh impossible unless you artificially upscale them you'll always end up with a slightly different image due to having more pixels which result in finer details which in turn change how the final image will look like. Post Processing glow is one very big issue here as Glow depends on every single pixel and its brightness according to your settings, having more pixels can mean more pixel are together which will have more glow in total or in worst case details, like glitter or sparkly reflections (sun on metal for instance) will be finer and thus occupy less pixel area compared to the smaller image, this results in glow potentially "vanishing" because there are not enough pixels bright enough and close together to fire off glow.

C) I do understand both RL and SL photography, in case you missed it i have been doing photography since 2007 (the very little that was.... "fancy") and i've been making a Viewer since 2011 that is made for photography. I'm sure i should be well prepared enough to criticize this. Besides, i'm not really criticizing anything, i'm saying that the bug report is basically a false report and should instead be a feature request to allow effects to scale with size which would most likely improve the chances of this ever happening. I don't see this ever being "fixed" as a bug. They will most likely just look into it and say the same i did.

Outside of the possibility there's a limit on blur radius, this is not a rendering issue, it's a user interface issue. I expect virtual reality to behave like actual reality because it's called... virtual reality, not because I'm naive about the technology. If there's naivete here, its on the part of whoever designed a DOF slider mechanism that doesn't factor in display resolution. If I can drag the DOF sliders to produce the same look at different resolutions, the algorithm can, too.

[Blarn Boo]

On 8/4/2021 at 9:49 PM, NiranV Dean said:

All screen effects requiring additional finetuning for higher resolutions is not a bug, this is normal behavior.

All your settings are made for the resolution you see on screen, which is the original window resolution, taking a snapshot at any other resolution requires finetuning these settings for that new resolution. In other words, you have to configure your settings for the target resolution. That's why there is a preview window.

All Viewers show this behavior and this is totally normal due to the way these effects are calculated (absolute rather than percentage based).

Note that higher resolutions require much higher/stronger settings which in turn need higher hardware requirements which multiply with the already increased hardware requirements of higher resolutions, this can and will quickly lead to a crash if your GPU isn't at the very least a GTX 1060 (6GB) and up. I tested up to 12.000 x 6.000 shots and even those will eventually lead to a crash if done often enough.

This is great! This clears it up for me and is in line with what I am experiencing.

I never used to push my resolution so high so I must just be experiencing this for the first time at this scale. Now I just wish the preview window was a little bigger so I could see the little artifacts before I hit save, haha.

Thanks!

 

 

[NiranV Dean]

22 hours ago, Madelaine McMasters said:

Outside of the possibility there's a limit on blur radius, this is not a rendering issue, it's a user interface issue. I expect virtual reality to behave like actual reality because it's called... virtual reality, not because I'm naive about the technology. If there's naivete here, its on the part of whoever designed a DOF slider mechanism that doesn't factor in display resolution. If I can drag the DOF sliders to produce the same look at different resolutions, the algorithm can, too.

As stated this is not just the case with Depth of Field "strength" this is the case with ALL shader effects that use an absolute size setting, that means ALL of them. SSAO, Depth of Field, Light Softening, heck even Antialiasing and the Snapshot filters too. In my Viewer almost every single additional feature does it too, Screen Space Reflections, Tone Mapping, Color Correction, Volumetric Lighting, Motion Blur... the only ones that don't are Vignette (because it has to scale with the image) and Lens Flare. Technically the  color posts (Sepia, Greyscale and Tonemapping/Color Correction) don't either because they just change colors per pixel but if they had some kind of area/size dependent algorythm they would too look completely different.