DynDOLOD 3.00 Alpha 196

[Phlunder]

Would it be possible for the Mesh Mask / Reference FormID field to allow multiple entries? Maybe its already a thing but I didn't find a way.

[sheson]

2 hours ago, Phlunder said:

Would it be possible for the Mesh Mask / Reference FormID field to allow multiple entries? Maybe its already a thing but I didn't find a way.

That is not possible. I suggest to create a rules file and use a text editor to copy/paste similar lines.

Use the feature that mesh masks only need to match a part of the path/filename.

[Phlunder]

10 minutes ago, sheson said:

That is not possible. I suggest to create a rules file and use a text editor to copy/paste similar lines.

Use the feature that mesh masks only need to match a part of the path/filename.

Yea I already have a text file to paste some of my mesh masks from. Found another way to filter out 2 meshes of a path that already has a rule applied, by creating a rule right above that acts like a filter.

[sheson]

On 10/29/2025 at 11:05 AM, mostwanted11 said:

It's an ESP I made. I will attach it.

Export: https://limewire.com/d/YNN9L#MaDhS90Z9C

NOTWL - Custom Replacers.esp 192.17 kB · 1 download

Get the current test version of DynDOLOD from https://dyndolod.info/Downloads/Test-Versions and generate just ultra tree LOD without grass LOD for Tamriel to see if there is anything different. Maybe make a test profile with just the necessary mods.

All seems in order comparing the debug logs between my current test version and your logs from Alpha 196.

However comparing your BTO and atlas texture on the left with the one I generated on the right shows a slight difference.

The trees I am comparing are:
 

According to the CRC32s reported by logs I have the same 3D tree LOD model and textures. So for now I just want to know if might notice a difference with the current test version or if I need to continue to look further.

[mostwanted11]

15 hours ago, sheson said:

Get the current test version of DynDOLOD from https://dyndolod.info/Downloads/Test-Versions and generate just ultra tree LOD without grass LOD for Tamriel to see if there is anything different. Maybe make a test profile with just the necessary mods.

All seems in order comparing the debug logs between my current test version and your logs from Alpha 196.

However comparing your BTO and atlas texture on the left with the one I generated on the right shows a slight difference.

The trees I am comparing are:
 

According to the CRC32s reported by logs I have the same 3D tree LOD model and textures. So for now I just want to know if might notice a difference with the current test version or if I need to continue to look further.

https://imgbox.com/g/zGQAAwK2FW

https://vimeo.com/manage/videos/1132434675
I don't see a difference personally. The one with grass is the previous output and the one without is the new output.

Here are the logs from this version: https://limewire.com/d/etTx5#zaHozyTVTM

I have another issue, it seems like Texgen generates grass lod textures seemingly randomly every time. For example, grass record groundcover.esp;00000819;t_grass_big02 sometimes gets flagged as ",false" and sometimes not. Sometimes a texgen output will have 315 grass lod files generated and sometimes it'll be 215. I have MinGrassModelVolume=0 and MinGrassModelHeight=0 in Texgen_SSE.ini because I think Texgen was ignoring the grass records of that mod because it's object bounds are weird? I recalculated them in CK and nothing changed(they were already calculated) until I set both of these to =0 but now they're being generated randomly I think.

The grass mod: https://www.nexusmods.com/skyrimspecialedition/mods/139896?tab=description

I generated 2 new Texgen Outputs right now using 196 and these are the results:

First one: https://justpaste.it/41ymd and its logs: https://limewire.com/d/54fqR#ZlIln1Z8gT

Second one: https://justpaste.it/4zy6a and its logs: https://limewire.com/d/59ViO#Y4riaVBo7D

I noticed this because one time I didn't have any tundra grass despite generating grass lods. I ended up saving the grass lod files from the texgen output that had the most amount after several tries.

Edited October 31 by mostwanted11

[sheson]

2 hours ago, mostwanted11 said:

https://imgbox.com/g/zGQAAwK2FW

https://vimeo.com/manage/videos/1132434675
I don't see a difference personally. The one with grass is the previous output and the one without is the new output.

Here are the logs from this version: https://limewire.com/d/etTx5#zaHozyTVTM

I have another issue, it seems like Texgen generates grass lod textures seemingly randomly every time. For example, grass record groundcover.esp;00000819;t_grass_big02 sometimes gets flagged as ",false" and sometimes not. Sometimes a texgen output will have 315 grass lod files generated and sometimes it'll be 215. I have MinGrassModelVolume=0 and MinGrassModelHeight=0 in Texgen_SSE.ini because I think Texgen was ignoring the grass records of that mod because it's object bounds are weird? I recalculated them in CK and nothing changed(they were already calculated) until I set both of these to =0 but now they're being generated randomly I think.

The grass mod: https://www.nexusmods.com/skyrimspecialedition/mods/139896?tab=description

I generated 2 new Texgen Outputs right now using 196 and these are the results:

First one: https://justpaste.it/41ymd and its logs: https://limewire.com/d/54fqR#ZlIln1Z8gT

Second one: https://justpaste.it/4zy6a and its logs: https://limewire.com/d/59ViO#Y4riaVBo7D

I noticed this because one time I didn't have any tundra grass despite generating grass lods. I ended up saving the grass lod files from the texgen output that had the most amount after several tries.

If you spend time creating a video, then make it obvious, which tree(s) fade into non existing branches. Verify those are actually LOD trees. Disable all post processing like ENB or CS.

As a test, half the alpha threshold in the used tree LOD model for a tree as reported by DynDOLOD_SSE_Object_LOD.txt. For example pruce_forest_medium01_summer.nif it is spruce_forest_medium01_summer_9961e054passthru_lod.nif for. Change alpha threshold form 110 to 55. Then generate a test LOD patch. Check and report if that has a visible effect.

TexGen reports it does not find enough opaque pixels, but it does not seem to be consistent for some reason. As example for me to check, upload Meshes\landscape\grass\grass_big02.nif and textures\Landscape\Grass\ng_c.dds

[mostwanted11]

1 hour ago, sheson said:

If you spend time creating a video, then make it obvious, which tree(s) fade into non existing branches. Verify those are actually LOD trees. Disable all post processing like ENB or CS.

As a test, half the alpha threshold in the used tree LOD model for a tree as reported by DynDOLOD_SSE_Object_LOD.txt. For example pruce_forest_medium01_summer.nif it is spruce_forest_medium01_summer_9961e054passthru_lod.nif for. Change alpha threshold form 110 to 55. Then generate a test LOD patch. Check and report if that has a visible effect.

TexGen reports it does not find enough opaque pixels, but it does not seem to be consistent for some reason. As example for me to check, upload Meshes\landscape\grass\grass_big02.nif and textures\Landscape\Grass\ng_c.dds

Apologies, I don't have the experience (and sometimes the time) to do such things, I thought I'd just point your attention towards this issue as you've fixed it before with that exe you uploaded in May, but I was forced to update and thus removing that exe. I'll provide the grass mesh and texture here although I've provided a link of the grass mod in case you want to further examine it's contents.

Good luck and sorry if I wasted your time.

grass_big02.nif ng_c.dds

Edited October 31 by mostwanted11

[sheson]

1 hour ago, mostwanted11 said:

Apologies, I don't have the experience (and sometimes the time) to do such things, I thought I'd just point your attention towards this issue as you've fixed it before with that exe you uploaded in May, but I was forced to update and thus removing that exe. I'll provide the grass mesh and texture here although I've provided a link of the grass mod in case you want to further examine it's contents.

Good luck and sorry if I wasted your time.

grass_big02.nif 4.46 kB · 1 download ng_c.dds 2.67 MB · 1 download

Back then the problem was crowns of LOD trees fading to nothing. https://stepmodifications.org/forum/topic/19903-dyndolod-300-alpha-196/page/678/#findComment-285705
As explained, the alpha threshold adjustment fix is included since Alpha-191. A test version is just a snapshot of the current version that is being worked on. An alpha version is just a snapshot of the same current version that has all known issues at the time addressed/tested and with the version number incremented.

Based on the provided BTO and atlas, the mipmaps seem to fade normally, so for now I am assuming this might be case of tree LOD assets that could use a lower alpha threshold to make the crowns a bit thicker. For the tests I suggest to disable ENB or CS to rule out post processing effects.

The provided debug logs report that the alpha threshold adjustment for the mipmap levels is applied. To test if the adjustment still works as it should, lower the alpha threshold of a corresponding tree LOD model and compare in the game that the crown of the LOD tree got thicker.

Can you double check if the test version also has an inconsistent number of grass LOD billboards being generated?

[mostwanted11]

1 hour ago, sheson said:

Back then the problem was crowns of LOD trees fading to nothing. https://stepmodifications.org/forum/topic/19903-dyndolod-300-alpha-196/page/678/#findComment-285705
As explained, the alpha threshold adjustment fix is included since Alpha-191. A test version is just a snapshot of the current version that is being worked on. An alpha version is just a snapshot of the same current version that has all known issues at the time addressed/tested and with the version number incremented.

Based on the provided BTO and atlas, the mipmaps seem to fade normally, so for now I am assuming this might be case of tree LOD assets that could use a lower alpha threshold to make the crowns a bit thicker. For the tests I suggest to disable ENB or CS to rule out post processing effects.

The provided debug logs report that the alpha threshold adjustment for the mipmap levels is applied. To test if the adjustment still works as it should, lower the alpha threshold of a corresponding tree LOD model and compare in the game that the crown of the LOD tree got thicker.

Can you double check if the test version also has an inconsistent number of grass LOD billboards being generated?

Yeah it might be an issue with that particular mod. I will test with another version that I think doesn't have that issue and report back.

As for the Grass, I don't think I can test a new output because it keeps telling me there's an issue with my drivers while generating the lods even though they're up to date. I had to completely trim my setup to the bare minimum in order to test the trees with that test version.

Edited October 31 by mostwanted11

[sheson]

12 hours ago, mostwanted11 said:

Yeah it might be an issue with that particular mod. I will test with another version that I think doesn't have that issue and report back.

As for the Grass, I don't think I can test a new output because it keeps telling me there's an issue with my drivers while generating the lods even though they're up to date. I had to completely trim my setup to the bare minimum in order to test the trees with that test version.

Change the alpha threshold in a specific LOD model to verify that the alpha threshold adjustment works when the texture is being added to the atlas. If different LOD models exist and they use the same full textures, compare their alpha thresholds and their shape names.

Upload the log, debug log and bugreport.txt (if it exists) from the test version. Copy and paste the error message with its Copy message to clipboard link as explain in https://dyndolod.info/Official-DynDOLOD-Support-Forum#Copy-and-Paste-Text

[Phlunder]

Sorry, already another mesh mask/rule question. I made Level 8 meshes for mountains, that still have the full texture assigned and differentiate between heavy and light snow coverage with the vertex alpha values. How can I make DynDOLOD recognize them with rules? Or is that only possible with a plugin? For example, instead of mountaincliff01_lod_1.nif I now have mountaincliff01_lod_1h.nif and mountaincliff01_lod_1l.nif and so on.

[sheson]

54 minutes ago, Phlunder said:

Sorry, already another mesh mask/rule question. I made Level 8 meshes for mountains, that still have the full texture assigned and differentiate between heavy and light snow coverage with the vertex alpha values. How can I make DynDOLOD recognize them with rules? Or is that only possible with a plugin? For example, instead of mountaincliff01_lod_1.nif I now have mountaincliff01_lod_1h.nif and mountaincliff01_lod_1l.nif and so on.

That can not be done with rules atm. You can try setting the LOD models on the base record, however without checking Prefer base record LOD assignments over rules (which might have unwanted side effects) it is probably not going to work as desired as long as there are LOD models with the "normal" filenames. The automatic filename matching is currently made to mimic the LOD of the vanilla game with limiting the use of l and h to LOD level 4.

I suggest to simply make a mesh mask rule that sets to use the Level0 LOD model for Level 8 instead for the desired mountains or to increase the LOD Level 4 distance. If there are custom Level0 LOD models installed that have such a high triangle count that it is noticeable performance wise with hardware form the past decade when they are also used in LOD Level 8, then I would propose those should be optimized regardless.

I might have a look to expand that automatic detection to the higher LOD levels in a future version.

[Phlunder]

9 minutes ago, sheson said:

That can not be done with rules atm. You can try setting the LOD models on the base record, however without checking Prefer base record LOD assignments over rules (which might have unwanted side effects) it is probably not going to work as desired as long as there are LOD models with the "normal" filenames. The automatic filename matching is currently made to mimic the LOD of the vanilla game with limiting the use of l and h to LOD level 4.

I suggest to simply make a mesh mask rule that sets to use the Level0 LOD model for Level 8 instead for the desired mountains or to increase the LOD Level 4 distance. If there are custom Level0 LOD models installed that have such a high triangle count that it is noticeable performance wise with hardware form the past decade when they are also used in LOD Level 8, then I would propose those should be optimized regardless.

I might have a look to expand that automatic detection to the higher LOD levels in a future version.

Thank you for the detailed explanation. I force Level0 mountain meshes in all LOD stages with a custom rule already, because I personally can't stand the vanilla Level1 and Level2 mountain meshes. Its not a performance concern, just wanted to err on the side of caution regarding optimization. 

Regarding automatic detection of LOD meshes, looking forward to future improvements. Its already impressive how much stuff is simply auto detected.

[luoxuangou]

Does the latest version of DynDOLOD still require xLODGen to generate terrain?

[sheson]

1 hour ago, luoxuangou said:

Does the latest version of DynDOLOD still require xLODGen to generate terrain?

https://dyndolod.info/Generation-Instructions#Prerequisites
Typically generate and install terrain LOD meshes and textures with xLODGen before generating the LOD mod with DynDOLOD, as the terrain LOD meshes are used to optimize object LOD meshes

https://dyndolod.info
Dynamic Distant Objects LOD (DynDOLOD) is a set of tools that is the advanced and easier version of xLODGen. The tools generate a comprehensive LOD mod for the entire load order of the different Skyrim and Enderal game versions for drastically enhanced and better matching tree LOD and object LOD plus optional dynamic LOD (hence the name), glow LOD, grass LOD, occlusion data and terrain underside in a few simple steps.

https://dyndolod.info/Help/xLODGen
DynDOLOD is the advanced and easier version of xLODGen for drastically improved tree LOD and object LOD generation plus many additional features for the Skyrim and Enderal versions.
Generate terrain LOD with xLODGen before using TexGen and DynDOLOD to generate drastically improved tree LOD and object LOD.