DynDOLOD 3.00 Alpha 198

[sheson]

1 hour ago, Dayraven said:

I have rerun the LOD, here are the new logs:

https://www.mediafire.com/file/xk55szp1e0c4dax/Logs2.rar/file

And some screenshots presenting the current state:

The 3D tree LOD model defines different textures for the leaves/crown than the full model. That should be corrected.

From the DynDOLOD_SSE_Object_Report.txt for this tree:

TreeAspen01 [TREE:0006A9E6] Meshes\landscape\trees\treeaspen01.nif [Original CRC32:4375A7E1] [Patched CRC32:42825682] using textures\landscape\trees\treeaspenbarkcomp.dds, textures\landscape\trees\treeaspenbarkcomp_n.dds, textures\mx\birch03_c.dds, textures\mx\birch03_n.dds

Level0: meshes\dyndolod\lod\trees\treeaspen01_4375a7e1passthru_lod.nif [CRC32:8BA32AED] using textures\dyndolod\lod\trees\treeaspen01_4375a7e1_trunk_1.dds, textures\dyndolod\lod\trees\treeaspen01_4375a7e1_trunk_1_n.dds, textures\dyndolod\lod\trees\treeaspen01_4375a7e1_trunk_2.dds, textures\dyndolod\lod\trees\treeaspen01_4375a7e1_trunk_2_n.dds, textures\mx\autumnal.dds, textures\mx\aspenbranch_n.dds

The 3D tree LOD model should define textures\mx\birch03_c.dds and textures\mx\birch03_n.dds instead.

[sheson]

3 hours ago, poiii said:

Hello,

Could you please take a look at the image I’ve attached and tell me whether the distant whiteness is related to LOD? Even after I increase the shadow distance in *SkyrimPrefs.ini*, the far-off terrain still looks washed out. It only appears normal once my character moves close enough for the proper shadows to load. I also wondered if EVLaS might be responsible, but it doesn’t seem to be.

https://ibb.co/zHHBRbf4

Please note that I am communicating with you through a translator.

Many thanks for your help!

Effects/shaders like shadows or lighting are not related to LOD. In the context of DynDOLOD, LOD generally relates to meshes and textures shown in the distances past the active cells.

See https://dyndolod.info/Official-DynDOLOD-Support-Forum#Rudimentary-Troubleshooting and https://dyndolod.info/How-LOD-Works

[z929669]

4 hours ago, poiii said:

Hello,

Could you please take a look at the image I’ve attached and tell me whether the distant whiteness is related to LOD? Even after I increase the shadow distance in *SkyrimPrefs.ini*, the far-off terrain still looks washed out. It only appears normal once my character moves close enough for the proper shadows to load. I also wondered if EVLaS might be responsible, but it doesn’t seem to be.

https://ibb.co/zHHBRbf4

Please note that I am communicating with you through a translator.

Many thanks for your help!

 

4 hours ago, sheson said:

The 3D tree LOD model defines different textures for the leaves/crown than the full model. That should be corrected.

From the DynDOLOD_SSE_Object_Report.txt for this tree:

TreeAspen01 [TREE:0006A9E6] Meshes\landscape\trees\treeaspen01.nif [Original CRC32:4375A7E1] [Patched CRC32:42825682] using textures\landscape\trees\treeaspenbarkcomp.dds, textures\landscape\trees\treeaspenbarkcomp_n.dds, textures\mx\birch03_c.dds, textures\mx\birch03_n.dds

Level0: meshes\dyndolod\lod\trees\treeaspen01_4375a7e1passthru_lod.nif [CRC32:8BA32AED] using textures\dyndolod\lod\trees\treeaspen01_4375a7e1_trunk_1.dds, textures\dyndolod\lod\trees\treeaspen01_4375a7e1_trunk_1_n.dds, textures\dyndolod\lod\trees\treeaspen01_4375a7e1_trunk_2.dds, textures\dyndolod\lod\trees\treeaspen01_4375a7e1_trunk_2_n.dds, textures\mx\autumnal.dds, textures\mx\aspenbranch_n.dds

The 3D tree LOD model should define textures\mx\birch03_c.dds and textures\mx\birch03_n.dds instead.

So I think this means that you are choosing "Autumnal Variety" in the Add-on, and "Autumnal" in the main mod. Choose the same type for both.

[sheson]

38 minutes ago, z929669 said:

So I think this means that you are choosing "Autumnal Variety" in the Add-on, and "Autumnal" in the main mod. Choose the same type for both.

Since the LOD model has the CRC32 of the full model, it means the textures in the LOD model need to be updated.

[z929669]

9 hours ago, sheson said:

Since the LOD model has the CRC32 of the full model, it means the textures in the LOD model need to be updated.

I actually just finished checking treeaspen01 LOD and full models. The AA add-on mod is the 'Performance' version with optimized crowns that combine the three shapes of the full model crowns. It also uses a smaller 256px branch texture than the full model's 4096px using a unique path name and custom mipmaps. In fact, all of the 'Performance' aspen LOD models (not just treeaspen01 LOD) use this smaller branch texture. The textures match, as can be seen in the following screens comparing freshly-generated LOD patches for the 'Performance' vs 'Quality' versions of the AA add-on (Autumnal). While the mipmaps of the Performance version may need another look, the color tones match (the LOD are a bit darker for both). Remember that the Quality version has LOD models using the exact crowns and texture paths as the full models, since they are direct copies of the crowns/textures and not optimized versions.

AA Performance Add-On (256) AA Performance Add-On (NiAP67) AA Performance Add-On (512) AA Performance Add-On (1024) AA Performance Add-On (2048) AA Quality Add-On

Performance logs

Quality logs

Q-P branch textures

@Dayraven

---

PS: The textures\mx\autumnal.dds texture has no mipmaps, because we leave it to DynDOLOD to create them during processing. This texture is only used by aspen LOD branches. The LOD branches are a bit too thin though, so either the 256px texture is too small to begin with, or NiAlphaProperty for branches needs to be lower in the LOD models (128 now just like the full models).

PSS: Changing NiAlphaProperty of the LOD model branches from 128 to 67 had no impact (see middle screenshot).

PSSS: Changing the texture resolution to 512px and 1024px makes a significant difference, so the starting size of 256px is definitely too small for my resolution (2160p). I assume it's because mipmap 0 at 256px is not high enough. NOTE: All variants of textures\mx\autumnal.dds here and in the Add-on were created from corresponding mipmaps of the original extures\mx\birch03_n.dds and saved as BC7 without mipmaps.

PSSSS: Finally, changing the resolution to 2048px yields slightly fuller LOD trees at this game distance (It's mid-distance LOD4). I presume that I will see no benefit at 4096px with my 2160p configuration, since this is the recommended texture clamp size at 4K. 

One additional consideration I hadn't considered until now is that I do set CrownBrightness=0.8 in my DynDOLOD_SSE.ini for better matching of EVT LODs, but that was for my old monitor and GFX card. I'm currently generating two additional compares of Performance and Quality at the default CrownBrightness=1 so that we can determine if there's really any issue that needs to be addressed.

... using default crownbrightness=1

LOD Reference AA-Performance (2048) AA-Quality

The Quality version seems like a very close match. The Performance version is a bit darker but also quite close. I attribute the difference to the fact that the optimized LOD models have slightly different geometry and fewer triangles rather than a texture difference.

textures

[sheson]

On 5/12/2025 at 9:32 AM, Dayraven said:

I have rerun the LOD, here are the new logs:

https://www.mediafire.com/file/xk55szp1e0c4dax/Logs2.rar/file

And some screenshots presenting the current state:

Since there was no reply yet, here are 3 variants for you to test. AA-3D-Perf-Tests.7z
Maybe you can report back which set has the best results in your case.

[z929669]

3 hours ago, sheson said:

Since there was no reply yet, here are 3 variants for you to test. AA-3D-Perf-Tests.7z
Maybe you can report back which set has the best results in your case.

I will also run these tests as well and let you know in the next day or two.

[z929669]

17 hours ago, sheson said:

Since there was no reply yet, here are 3 variants for you to test. AA-3D-Perf-Tests.7z
Maybe you can report back which set has the best results in your case.

I've completed the tests. I also added two tests to account for the sparse aspen LOD foliage that I discovered was caused by the custom LOD textures having too low resolution at 256px, presumably due to my 4K configuration (2160p). The AA optimized 'Performance' LODs were originally tested under 1080p-1440p resolution and DynDOLOD settings. At the time, we had a good color-tone match between full and LOD. The shape match was satisfactory as well.

As mentioned previously, the custom 256px LOD textures were created by extracting the corresponding mipmap from the original textures and saved without mipmaps. All I did for the additional two tests was extract the 2046px mipmaps from the original AA textures, saving again without mipmaps. These are only utilized by the LOD models, hence the different texture paths from the full models.

The baseline test is 'default'. It's just AA Autumnal Variety with 'big' trees option + AA Add-On for this flavor without any of your changes.

The effect of using the higher-resolution custom LOD textures follow, so that we can examine more apples-to-apples comparison for your tests, given we see the same color tones with the different branch texture resolutions:

LOD Reference

#1 default #2 default-2048 #3 normals #4 normals-2048

Now to the test compares. I'm using the '2048' variants in the following so that we don't need to 'unsee' the sparse crowns in LOD in the respective compares that use the custom LOD textures (default & normals tests).

#2 default-2048 #4 normals-2048 #5 normals-textures #6 normals-textures-effect

While there are some subtle differences, the color-tone mismatch remains in all cases.

LOGS

• 1_default
• 2_default-2048
• 3_normals
• 4_normals-2048
• 5_normals-textures
• 6_normals-textures-effect

@Dayraven

---

PS: I see all the changes you made to the meshes, but how exactly did you change the normals, tangents, and bitangents for each vertex? I couldn't reproduce that in NifSkope using the tools I'm aware of.

[sheson]

5 hours ago, z929669 said:

I've completed the tests. I also added two tests to account for the sparse aspen LOD foliage that I discovered was caused by the custom LOD textures having too low resolution at 256px, presumably due to my 4K configuration (2160p). The AA optimized 'Performance' LODs were originally tested under 1080p-1440p resolution and DynDOLOD settings. At the time, we had a good color-tone match between full and LOD. The shape match was satisfactory as well.

As mentioned previously, the custom 256px LOD textures were created by extracting the corresponding mipmap from the original textures and saved without mipmaps. All I did for the additional two tests was extract the 2046px mipmaps from the original AA textures, saving again without mipmaps. These are only utilized by the LOD models, hence the different texture paths from the full models.

The baseline test is 'default'. It's just AA Autumnal Variety with 'big' trees option + AA Add-On for this flavor without any of your changes.

The effect of using the higher-resolution custom LOD textures follow, so that we can examine more apples-to-apples comparison for your tests, given we see the same color tones with the different branch texture resolutions:

LOD Reference

#1 default #2 default-2048 #3 normals #4 normals-2048

Now to the test compares. I'm using the '2048' variants in the following so that we don't need to 'unsee' the sparse crowns in LOD in the respective compares that use the custom LOD textures (default & normals tests).

#2 default-2048 #4 normals-2048 #5 normals-textures #6 normals-textures-effect

While there are some subtle differences, the color-tone mismatch remains in all cases.

LOGS

• 1_default
• 2_default-2048
• 3_normals
• 4_normals-2048
• 5_normals-textures
• 6_normals-textures-effect

@Dayraven

---

PS: I see all the changes you made to the meshes, but how exactly did you change the normals, tangents, and bitangents for each vertex? I couldn't reproduce that in NifSkope using the tools I'm aware of.

I suggest to do these test in the vanilla game with just the mods you are testing. In particular without ENB, CS, weather mods. Something seems odd in the screenshots.

The steps were face normals, smooth normals with 120, update tangents.

I suggest to define the full textures in the assets and simply let DynDOLOD handle the resizing onto the object LOD texture atlas instead of creating smaller "full" textures that is resized by DynDOLOD onto the atlas just the same.

https://dyndolod.info/Help/Advanced-Mode
The first time DynDOLOD is started, some settings like the Max tile size LOD, Max tile size full and the Max tile size billboard are adjusted to the current game resolution as determined by the Prefs, Custom or Display Tweaks INI.

Typically 256 for 1080p, 512 in your case.

[z929669]

16 hours ago, sheson said:

I suggest to do these test in the vanilla game with just the mods you are testing. In particular without ENB, CS, weather mods. Something seems odd in the screenshots.

The steps were face normals, smooth normals with 120, update tangents.

I suggest to define the full textures in the assets and simply let DynDOLOD handle the resizing onto the object LOD texture atlas instead of creating smaller "full" textures that is resized by DynDOLOD onto the atlas just the same.

https://dyndolod.info/Help/Advanced-Mode
The first time DynDOLOD is started, some settings like the Max tile size LOD, Max tile size full and the Max tile size billboard are adjusted to the current game resolution as determined by the Prefs, Custom or Display Tweaks INI.

Typically 256 for 1080p, 512 in your case.

I repeated the tests under a barebones modded vanilla, so vanilla weather and no post processing. The results are basically the same. As with the previous tests, the largest improvement is the revised normals; although, both using the default textures and higher Lighting Effect 1 very slightly raised the LOD crown brightness.

I've also included a test with the 'Quality' version of the add-on. The result of that is essentially the same as I showed in my post just before my previous test post. This has an obviously better match with full trees and demonstrates that using the full crowns without optimization yields the best result. The only issue is that these crowns are too complex for LOD.

LOD Reference

1_default-2048 2_normals-2048 3_normals-textures 4_normals-textures-effect

5_default-Quality

LOGS

• 1_default-2048
• 2_normals-2048
• 3_normals-textures
• 4_normals-textures-effect
• 5_default-Quality

[dionysist]

I am experiencing missing windmill blades, even in the active cell.  If I disable DynDOLOD, the blades reappear.

I use the 'High' preset.

Bugreport.txt: https://mega.nz/file/6mAWhCKZ#XrvvcrikQR0HH5bXu8K79cDwX66WRkP2vXD-KEwGbzc

DynDOLOD_TES5VR_Dynamic_LOD.txt: https://mega.nz/file/3zJHFZ7J#acnnDaZdB-lYBlKEy6rj0Q6XzTrxgyQcxEKJWEonf6k

DynDOLOD_TES5VR_log.txt: https://mega.nz/file/yiowGJaS#azCYM79FlC3yx48f9z2HbnQKmGOW29PGTA1eXKp2OtA

DynDOLOD_TES5VR_TexturesUsed_Tamriel.txt: https://mega.nz/file/vrhnzIyR#l2joloQrf88vDceVOOIq9D0E4_dSFg1gjqAhPgUetGs

TexGen_TES5VR_Debug_log.txt: https://mega.nz/file/inw3BSJQ#Kf7TKzHthfj6D1uhgtL9pRhK1fKDtHl9QDucnfHQh3I

TexGen_TES5VR_log.txt: https://mega.nz/file/DnwjCKrZ#f9Msb8mHoPR3lUAztnEUQycPzQsMSfE1u_Z-PO6Pvtk

default farmhousewindmillfan rules:

 

Please let me know if there is any other information needed.

[sheson]

4 hours ago, dionysist said:

I am experiencing missing windmill blades, even in the active cell.  If I disable DynDOLOD, the blades reappear.

I use the 'High' preset.

Bugreport.txt: https://mega.nz/file/6mAWhCKZ#XrvvcrikQR0HH5bXu8K79cDwX66WRkP2vXD-KEwGbzc

DynDOLOD_TES5VR_Dynamic_LOD.txt: https://mega.nz/file/3zJHFZ7J#acnnDaZdB-lYBlKEy6rj0Q6XzTrxgyQcxEKJWEonf6k

DynDOLOD_TES5VR_log.txt: https://mega.nz/file/yiowGJaS#azCYM79FlC3yx48f9z2HbnQKmGOW29PGTA1eXKp2OtA

DynDOLOD_TES5VR_TexturesUsed_Tamriel.txt: https://mega.nz/file/vrhnzIyR#l2joloQrf88vDceVOOIq9D0E4_dSFg1gjqAhPgUetGs

TexGen_TES5VR_Debug_log.txt: https://mega.nz/file/inw3BSJQ#Kf7TKzHthfj6D1uhgtL9pRhK1fKDtHl9QDucnfHQh3I

TexGen_TES5VR_log.txt: https://mega.nz/file/DnwjCKrZ#f9Msb8mHoPR3lUAztnEUQycPzQsMSfE1u_Z-PO6Pvtk

default farmhousewindmillfan rules:

Please let me know if there is any other information needed.

Read the first post and/or https://dyndolod.info/Official-DynDOLOD-Support-Forum#Post-Logs which DynDOLOD debug log to also upload when making posts. Also upload the papyrus.log and DynDOLOD.log from DynDOLOD.DLL as explained in https://dyndolod.info/Official-DynDOLOD-Support-Forum#Post-Logs

Make a useful screenshot with more informative console of the full model.

https://dyndolod.info/Official-DynDOLOD-Support-Forum#In-Game-Screenshots
When making screenshots of full models, open console and click the object in question to show its reference form ID. Note that multi pass snow shaders can prevent full models from being selected. Temporarily disable any mod that prevents getting the form ID of full models in case getting the information is not possible otherwise.
If making screenshots of objects missing with DynDOLOD output enabled, make a screenshot of the objects as explained above without DynDOLOD output enabled.

https://dyndolod.info/Official-DynDOLOD-Support-Forum#Copy-and-Paste-Text
Do not post screenshots of the tools, unless it is about visual issues with the user interface. If necessary, convey the information in text form. The debug log - as explained below - typically contains most of the selected settings and options already

[dionysist]

12 hours ago, sheson said:

Read the first post and/or https://dyndolod.info/Official-DynDOLOD-Support-Forum#Post-Logs which DynDOLOD debug log to also upload when making posts. Also upload the papyrus.log and DynDOLOD.log from DynDOLOD.DLL as explained in https://dyndolod.info/Official-DynDOLOD-Support-Forum#Post-Logs

Make a useful screenshot with more informative console of the full model.

https://dyndolod.info/Official-DynDOLOD-Support-Forum#In-Game-Screenshots
When making screenshots of full models, open console and click the object in question to show its reference form ID. Note that multi pass snow shaders can prevent full models from being selected. Temporarily disable any mod that prevents getting the form ID of full models in case getting the information is not possible otherwise.
If making screenshots of objects missing with DynDOLOD output enabled, make a screenshot of the objects as explained above without DynDOLOD output enabled.

https://dyndolod.info/Official-DynDOLOD-Support-Forum#Copy-and-Paste-Text
Do not post screenshots of the tools, unless it is about visual issues with the user interface. If necessary, convey the information in text form. The debug log - as explained below - typically contains most of the selected settings and options already

Papyrus log: https://pastebin.com/jcpH6s92

Truncated DynDOLOD_TES5VR_log.txt: https://pastebin.com/m6CUzt7w

The entire DynDOLOD.log: https://pastebin.com/G1hfQBY7

It's quite nearly impossible to use the console in VR to select models and take a screenshot. The model, though, is FarmhouseWindMilFan.nif and ref form ID is 00096272, base form ID is 000FC611

[sheson]

10 minutes ago, dionysist said:

Papyrus log: https://pastebin.com/jcpH6s92

Truncated DynDOLOD_TES5VR_log.txt: https://pastebin.com/m6CUzt7w

The entire DynDOLOD.log: https://pastebin.com/G1hfQBY7

It's quite nearly impossible to use the console in VR to select models and take a screenshot. The model, though, is FarmhouseWindMilFan.nif and ref form ID is 00096272, base form ID is 000FC611

Read the first post and/or https://dyndolod.info/Official-DynDOLOD-Support-Forum#Post-Logs which entire DynDOLOD debug log to also upload when making posts.

You provided a DynDOLOD log, which shows DynDOLOD being started and then closed without generating. I hope that doesn't mean the debug log form the generation did get replaced by that as well. For better troubleshooting we need the debug log that shows the generation.

Are you guessing the reference form ID of the original windmill fan?
See if https://www.nexusmods.com/skyrimspecialedition/mods/140752 helps.
Use prid 00096272 or whatever the ref ID is to manually "select" the reference, then disable/enable to see/verify that you got the right one.
Alternatively, after clicking an object , you can try using the scrollwheel to cycle through refs. You can also try to use tcl to walk really close to the reference before clicking.

How do you start/load the game to reproduce the problem?
Does fast traveling to Solitude for example and back make a difference?

[SnootyCape]

Hi,

I've been getting a consistent error when it comes to installing DynDolod Output after I run DynDolod using MO2.

No matter how many times I generate it, I'll get a CRC error when it comes to installing, usually at different percentages for each Lod I generate.

Any ideas on why this is happening? Any help would be appreciated.