GUIDE [WIP] DDSopt & Texture Overhauls

[z929669]

Discussion thread:
DDSopt Guide by STEP
Wiki Link

---

 

GET DDSopt:
Github Pre-release versions
Official Nexus versions (select pre-release update 4)

[z929669]

The DDSopt guide now has an updated version of the HRDLC repair procedure using improved batch files that don't use drive substitution. The revised procedure should be much more robust.

Thanks Kelmych. I'll check it out.

Works great. I added a few missing placeholder.txt files that were iin dlc01.

[EssArrBee]

So, I'm still kinda fuzzy on the packing of the optimized files back into a bsa. Do we use BSAopt because it is so much better, or what? What exactly is the difference between the two when it comes to archives?

[z929669]

Optimising textures is something I never really understood - talking about mipmaps' date=' normal maps and all the DDSOpt settings malarchy... I would really appreciate if someone explained to me what the settings in DDSOpt should be to optimise, say, Skyrim Flora Overhaul to 1K with half normals (as that what the common pratice seems to be).

 

I plan to optimise the main exterior texture replacers (namely Skyrim HD and SFO + Automatic Variants), but I am not sure how to do it, and reading through 101 pages of this thread isn't something I would be looking forward to. :)

 

Can anyone help?[/quote']

Unfortunately there haven't been a lot of comparison tests reported, especially with changing formats and resolutions for normal maps, to allow making good recommendations for anything other than architectural textures and some landscape textures for large objects like mountains. You could read through the entire 100+ pages of this thread and find relatively little in definitive test results and only a few useful anecdotal findings. It's particularly hard to make any recommendations for SFO since the author has already done some optimization, and many of the textures are rectangular (which complicates optimization). As I mentioned in a recent post, I'm working on some new processing steps, initially for the vanilla textures, to greatly simplify the optimization process. The problem I have isn't on creating some batch processing steps to eliminate a number of the current tedious steps - that's fairly straightforward - but on coming up with some default parameter recommendations especially for how to handle normal maps (e.g., does it actually make sense to use additional compression/lower resolution for normal maps with the vanilla textures since so few are architectural or landscape related; and is the answer different for a system running extreme STEP vs. a lower end system).

The OP condenses most of the userful info in this thread as doe sthe DDSopt guide. I just finished a detailed optimization of all vanilla textures. I treated terrain normals and MSN as special cases. Attached are my logs (named according to treatment). I think that it makes good sense to constrain all vanilla normals to 1k and to ensure that the mip algorithms be applied for all using moderate settings. I omit the foliage mip algorithm for all but vanilla textures though. The gamma correction and contrast enhancements are key though.

So' date=' I'm still kinda fuzzy on the packing of the optimized files back into a bsa. Do we use BSAopt because it is so much better, or what? What exactly is the difference between the two when it comes to archives?[/quote']

No difference at all. BSAopt is just simpler to use. I don't use it at all myself.

Optimization.7z

[hellanios]

Should we still use the HighResTexturePack DLC Tweaks mod?

[WilliamImm]

I don't think so. The Unofficial High Res Fixpack includes most, if not all of the fixes from the HighRes tweaks.

[Kelmych]

That's also what z and I concluded; any remaining fixes are replaced by other textures from a variety of mods.

[Kelmych]

The OP condenses most of the userful info in this thread as doe sthe DDSopt guide. I just finished a detailed optimization of all vanilla textures. I treated terrain normals and MSN as special cases. Attached are my logs (named according to treatment). I think that it makes good sense to constrain all vanilla normals to 1k and to ensure that the mip algorithms be applied for all using moderate settings. I omit the foliage mip algorithm for all but vanilla textures though. The gamma correction and contrast enhancements are key though.

It looks like the diffuse terrain didn't get remipped (5093 skipped textures), by the way. Did you manually select the different categories you used? When you finished how did you repackage the data into archive files? Are you planning to do the DLC textures also?

[z929669]

The OP condenses most of the userful info in this thread as doe sthe DDSopt guide. I just finished a detailed optimization of all vanilla textures. I treated terrain normals and MSN as special cases. Attached are my logs (named according to treatment). I think that it makes good sense to constrain all vanilla normals to 1k and to ensure that the mip algorithms be applied for all using moderate settings. I omit the foliage mip algorithm for all but vanilla textures though. The gamma correction and contrast enhancements are key though.

It looks like the diffuse terrain didn't get remipped (5093 skipped textures), by the way. Did you manually select the different categories you used? When you finished how did you repackage the data into archive files? Are you planning to do the DLC textures also?

I processed the diffuse terrain along with all the rest in my final run, so they passed the verification I assume. I forgot the PNG though, so I'll need to copy those over as well as the interface textures that make sense. It may be a good idea to put anything like these in the DDSopt INI so that you don't need to treat them in your BATs.

 

I have yet to test everything and confirm the results I expect.

 

I ran as many runs as you see logs using filters applied manually. Just repackaged into 7z fast/non-solid. Still have not messed with the DLC other than HRDLC.

[Gopher]

I may be imagining this but after I optimize the textures the blocky faces problem seems to be back. I know it never really 'went' but was reduced by the HR pack, but I seem to notice it more now. Is it because I am looking for problems or did I do something wrong? :)

 

EDIT: Nevermind, I had not set the compression correctly for the normal maps.

[hellanios]

Hey guys is there a changelog anywhere or can you create one if it doesn't exist?

[Kelmych]

Hey guys is there a changelog anywhere or can you create one if it doesn't exist?

The guide is on a wiki so under Page Tools at the top select History to show when it was changed and to look at previous versions. Some of the changes are annotated. Is there a particular set of changes you were interested in? Right now we are making some changes in the vanilla texture optimization steps in the DDSopt Optimization tab. There were comments that the procedure was a bit tedious and error prone. We have also been looking at a better set of resolution/format parameters in the DDSopt Constraints tab use with for normal map texture files. This should be finished very soon.

[hellanios]

Hey guys is there a changelog anywhere or can you create one if it doesn't exist?

The guide is on a wiki so under Page Tools at the top select History to show when it was changed and to look at previous versions. Some of the changes are annotated. Is there a particular set of changes you were interested in? Right now we are making some changes in the vanilla texture optimization steps in the DDSopt Optimization tab. There were comments that the procedure was a bit tedious and error prone. We have also been looking at a better set of resolution/format parameters for normal map texture files. This should be finished very soon.

No, I just couldn't fing any changelog. Anyway, thanks for the info, man. You guys rock!!

[Kelmych]

 

The OP condenses most of the userful info in this thread as doe sthe DDSopt guide. I just finished a detailed optimization of all vanilla textures. I treated terrain normals and MSN as special cases. Attached are my logs (named according to treatment). I think that it makes good sense to constrain all vanilla normals to 1k and to ensure that the mip algorithms be applied for all using moderate settings. I omit the foliage mip algorithm for all but vanilla textures though. The gamma correction and contrast enhancements are key though.

It looks like the diffuse terrain didn't get remipped (5093 skipped textures), by the way. Did you manually select the different categories you used? When you finished how did you repackage the data into archive files? Are you planning to do the DLC textures also?

I processed the diffuse terrain along with all the rest in my final run, so they passed the verification I assume. I forgot the PNG though, so I'll need to copy those over as well as the interface textures that make sense. It may be a good idea to put anything like these in the DDSopt INI so that you don't need to treat them in your BATs.

 

I have yet to test everything and confirm the results I expect.

 

I ran as many runs as you see logs using filters applied manually. Just repackaged into 7z fast/non-solid. Still have not messed with the DLC other than HRDLC.

Part of the reason for the new batch file is to automatically take care of any direct copies like the .png files so it doesn't require a separate set of user steps. It also eliminates the need for special handling to eliminate any text and other non-texture files since they are never copied from the extracted BSAs to the folders that DDSopt itself uses; DDSopt runs on some temporary folders that contain only texture files and it doesn't run directly on the Vanilla Extracted folder.

 

Which interface textures should be copied over directly, or is it best to just put them in ddsopt.ini; I expect the latter is the better approach for these if they are .dds files?

[Kelmych]

The new batch file seems to work well in my (limited) testing. The DDSopt Optimization tab of the DDSopt guide has been modified to describe the use of the batch file, which resulted in eliminating the steps where the DDSopt GUI was used to select and deselect portions of the textures. Screenshots have also been added to this section to make it easier to determine the settings to use in DDSopt while optimizing different types of textures. The Quickstart tab was also modified for consistency with this approach.

 

One of the purposes of the simplification was to reduce the errors that were happening too frequently as users were following the at-times-intricate steps in the DDSopt guide. We would appreciate comments on whether the changes help, and what issues remain that we are capable of addressing (the just-put-the-BSAs-in-a-folder-and-push-one-button-and-poof-everything-is-optimized is still a bit out of reach).

 

Currently the optimized output creates one archive file for each of the vanilla BSA files. Is this what DDSopt users prefer? I noticed, for example, that z's example above combined all of the HRDLC into a single file vs. 3 separate ones. Do users prefer to have some or all of the optimized normal maps in different archive files or integrated with ordinary textures (as it is in the current DDSopt guide)? I've planning to provide a small batch file that uses 7zip to separately archive each of the optimized loose file folders, but I'd prefer to wait until we have a better idea of how we want to organize the output.

[hellanios]

Comment edited.