Open hexabits opened 1 year ago
All files in all Meshes BA2 reading correctly. This doesn't mean much until .mesh unknowns are figured out. It is fully readable but there are buffers with unknown uses.
I have .mesh loading rudimentarily into a new Shape class, but we're still figuring out .mesh unknowns.
I also managed to decode the geometry. but I still don’t understand how TBNs are encrypted. And the order of reading the data is not the same as the order of reading the vertex data
in 4 bytes the vectors x y z are transmitted.
as I understand it, these are vectors converted to RGBA
and so, I managed to extract TBN using the recent Zenimax TES Online project. TBN in File.mesh are shaded in a two-dimensional coordinate system in the same way as it is done in GR2 2byte per vector.
00 40 00 00 02 CC 00 C0 23 81 00 C0 (Half_Float Type) just in File.mesh it is encrypted in the Short_Signed type which needs to be converted to Float
I really don’t have any idea about the two-dimensional calculation method, maybe someone knows how to convert 3 into 2 vectors or how it works in general?
Doing God's work as per usual fellas and it is much appreciated!
Any ballpark idea on when we might have an updated NifSkope to open these assets?
and so, I managed to extract TBN using the recent Zenimax TES Online project. TBN in File.mesh are shaded in a two-dimensional coordinate system in the same way as it is done in GR2 2byte per vector. 00 40 00 00 02 CC 00 C0 23 81 00 C0 (Half_Float Type) just in File.mesh it is encrypted in the Short_Signed type which needs to be converted to Float
I really don’t have any idea about the two-dimensional calculation method, maybe someone knows how to convert 3 into 2 vectors or how it works in general?
TBN are R10G10B10A2, the sign of the bitangent is two of the bits in the leftover 4 bits between the two buffers.
Anyway, I've been slowly pushing things to my branch. I haven't pushed the new Shape class for rendering .mesh, I'm still working on gltf import/export.
https://github.com/fo76utils/ce2utils/blob/main/src/meshfile.cpp
As of this comment this decoding is still incomplete FYI. I provided them the full .mesh but it doesn't look to be implemented in their code yet.
и вот мне удалось извлечь TBN с помощью недавнего проекта Zenimax TES Online. TBN в File.mesh заштрихованы в двумерной системе координат так же, как это сделано в GR2 по 2 байта на вектор. 00 40 00 00 02 CC 00 C0 23 81 00 C0 (Тип Half_Float) просто в File.mesh он зашифрован в типе Short_Signed, который нужно преобразовать в Float
Я правда понятия не имею о двумерном методе расчета, может быть кто-нибудь знает, как преобразовать 3 в 2 вектора или как это вообще работает?
TBN — это R10G10B10A2, знак битангенса — это два бита в оставшихся 4 битах между двумя буферами.
В любом случае, я медленно продвигал вещи в свою ветку. Я еще не внедрил новый класс Shape для рендеринга .mesh, я все еще работаю над импортом/экспортом gltf.
how to convert RGBA to float for vector values?
OK I have basic gltf export working with the NiNode structure intact (allows transforming meshes without affecting their vertex data, which then eases re-import):
But I may still take a day or two to sort out:
OK I have basic gltf export working with the NiNode structure intact (allows transforming meshes without affecting their vertex data, which then eases re-import):
But I may still take a day or two to sort out:
- Mesh LODs. I can export these as separate meshes in gltf, but I realized that for large NIFs, all the LOD would be visible on import and look very ugly, and be annoying to hide every non-LOD0 mesh. I need to figure out what is best for user workflows while still exporting LOD.
- Weights and joints for skinned meshes. I should have all the data necessary to do so, I just haven't tackled it yet.
- Using gltf extras to encompass arbitrary NIF data, theoretically a NIF could be entirely encapsulated by the gltf format. For example, the gltf could contain data to recreate arbitrary NiObjects like BSXFlags on gltf import. This would mean gltf could create fully functional NIFs when importing into NifSkope.
- Importing. I only have this half implemented at the moment.
Would you be willing to share this verion at all? I know it's not fully functional for proper use, but if you've got nif files exporting/importing to Blender thats all some of us need to get started playing around with textures in Substance Painter.
Thanks for all the work either way.
Would you be willing to share this verion at all? I know it's not fully functional for proper use, but if you've got nif files exporting/importing to Blender thats all some of us need to get started playing around with textures in Substance Painter.
Thanks for all the work either way.
@Caseter
I understand the feeling believe me but the longer NifSkope goes without a second update after this hypothetical update, the more people are exporting gltf that aren't fully complete (such as the LOD and gltf extras I mentioned) and the less complete the data will be on import back into NIF. So I'm trying to shove as much data into gltf using gltf extras as I can so that people don't have to keep re-exporting to gltf to get more complete data later.
I will try to get a release out within the next 48 hours regardless, I know this has been a long process and I apologize. Last weekend I was on vacation for several days and then I've been slow on progress this week so I know people are getting antsy.
Also I haven't yet integrated DirectXMesh lib so the end of .mesh can't be generated yet, I am doing that as soon as I make export a little more complete. If it ends up taking too long, I'm just going to not allow import for the initial release.
Would you be willing to share this verion at all? I know it's not fully functional for proper use, but if you've got nif files exporting/importing to Blender thats all some of us need to get started playing around with textures in Substance Painter. Thanks for all the work either way.
@Caseter
I understand the feeling believe me but the longer NifSkope goes without a second update after this hypothetical update, the more people are exporting gltf that aren't fully complete (such as the LOD and gltf extras I mentioned) and the less complete the data will be on import back into NIF. So I'm trying to shove as much data into gltf using gltf extras as I can so that people don't have to keep re-exporting to gltf to get more complete data later.
I will try to get a release out within the next 48 hours regardless, I know this has been a long process and I apologize. Last weekend I was on vacation for several days and then I've been slow on progress this week so I know people are getting antsy.
Also I haven't yet integrated DirectXMesh lib so the end of .mesh can't be generated yet, I am doing that as soon as I make export a little more complete. If it ends up taking too long, I'm just going to not allow import for the initial release.
Is your accomplishment good enough to export X mesh as a stl? As i really want to have a few models/meshes 3d printed. And simply just a complete model as vertices is enough for me :P
I've sorted all the skinned mesh issues on export, so I'm trying to prepare a release. The skinning had issue after issue and took longer than expected so I've decided to release with export-only for now and work on import after initial release.
https://github.com/niftools/nifskope/assets/170077/74511fe9-884a-4fd4-bcff-53939baf8f9b
https://github.com/niftools/nifskope/assets/170077/676c20b3-b655-4dcb-b7cb-8fca4df9a1c8
https://github.com/niftools/nifskope/assets/170077/53e8f0d0-e6cf-424d-9241-c697fe1f1a00
Also, the later two videos show an LOD helper script written by Inaki for this.
Also I will document this with a guide/etc. but FYI stuff rigged on the human actors will need to copy/paste the COM branch of the human skeleton.nif into the mesh NIF before export. At least for the time being until I automate this. The creatures have a COM_Twin skeleton inside them and do not need this step.
I've put up a very early release, please read README_GLTF at least. https://github.com/hexabits/nifskope/releases
For the ranger badge, how did beth deal with that one? Does it have a complete model somewhere? Or is it literally just bumpmap textures that makes it 3d in-game? And do you guys know the technique to align the normals properly to make out the rest of the "model" that can be made solid and export as stl for instance? O tried all sorts of unwrappings yesterday, but neither made it "complete"
Thank you, sir, for your hard work. This is fantastic! Finally, we can view and export models in Nifskope, which is of great significance for MOD authors in texture redesign tasks. It's great. However, I'd like to inquire if there is currently a way to modify the models. Simply modifying textures has become tiresome for people. We urgently need a convenient method to modify models in the game and export them to the NIF format with .mesh files in the correctly named folders, facilitating model replacement work for mod authors. I believe this is what players and mod authors all around the world have been eagerly anticipating day and night. Thanks again for your efforts, and I look forward to seeing a version of Nifskope soon that allows modification and import/export of models.
Hello! First I would like to thank everyone who has and is currently working on Nifskope for all of your hard work. You guys have made it possible for me to finally put my foot in the door for a career in something I truly enjoy. Secondly I would like to state that I am having issues with the current version of Nifskope and I apologize if this has been stated before but I'm not familiar with the way this website works. For Starfield it seems that about 75% of the meshes are visible when I open the file but the files that I'm looking to edit are seemingly invisible. When I open the naked_f.nif everything seems to be working correctly, but when I open the naked_m.nif nothing shows up. Is this issue something I am doing wrong or is this something the developers are currently working to fix?
Hi there, Sorry if i write here but discord channel invitation is not working anymore, and the readme seciton of troubleshooting is empty... I've downloaded the software, then I added Starfield data folder, but when I try to add archives they are just being ignored (while fallout 4 ones are all listed visible) and when I extracted the geometry folder and put it in data, loading nif files still gives issues. What am I doing wrong? Thanks for attention and for the modding tools guys!
---EDIT----
Problem solved. I'm an idiot. I downloaded the latest version, then unzipped the old one. Sorry for bothering you, I'll keep the message just in case there is some other genius like me around. And I kinda hope so... would make me feel better not being alone :/
I'm having problem with importing outfits upperbodies I have followed the instruction of copying Skeleton NiNode and paste It on the cloth model... but upperbodies meshes refuse to be imported in Blender.
I'm using Blender 3.4
I tried to tweak import settings but no luck. Help?
So, I did a few test before answering. Ecliptig legs, they work. Marine armor plates addon, it works. UC Navy pouches addon, it works. Constellation body, it works. UC police works. Neon security body, it doesn't work. Try with some other meshes just to make sure your process is right. The above mentioned, for instance: if they work for me, they should work for you too. I'm on blender 3.6
@Coverop @GabGone Sorry, I've covered it on multiple Discord servers but failed to really mention it on GitHub, as I found out about a while after release. Currently stuff with Havok Cloth bones doesn't export correctly. These bones do not exist in the COM skeleton because they likely get added by Havok at runtime.
The only info I have is the bone index/weight plus the inverse bone matrix for the Havok bone which isn't enough to reconstruct it in the full armature in Blender. I am currently finding the best way to fail on these meshes still. They will simply render incorrectly in Blender because I can't put them in the correct place in the armature, but the underlying data should all be fine for re-import.
Also, there was a secondary issue with Bethesda including close to 0.0 vertex weights where they should have been 0.0, and I've also figured this out now. Everyone should just not import skinned meshes until I update the release. Due to the way glTF works with joints and weights, this ends up messing up a few of the weights in the first vertex group, which was also the reason Guess Original Bind Pose had to be unchecked. You should no longer have to do this after I update the release.
Thanks for the answer. I've been looking for a discord page about nifskope, but I've only found expired invitations. Would you please post a working link?
Thanks for the answer. I've been looking for a discord page about nifskope, but I've only found expired invitations. Would you please post a working link?
There is no official server anymore, I closed it years ago. I did remove the stuff that no longer exists from the app and readmes. Did I miss something somewhere? (I'm aware the old niftools page still has a Discord widget though)
I have updated the release (same link): https://github.com/hexabits/nifskope/releases/tag/v2.0.dev9
Please read the 2023-09-29 notes.
Hello! I am having a small issue, when i export a model into nifskope it appears as invisible? and when i export it into blender it only is the bones, its likely a problem on my end but i wanted to see if anyone else had this problem?
Latest version was working fine on me. Can you please tell us which niff is giving you the issue? So far, I can confirm we're able to:
Latest version was working fine on me. Can you please tell us which niff is giving you the issue? So far, I can confirm we're able to:
- copy branches of niff to make mash ups, but it seems that you can't move/scale branches (still testing) on armors
- copy branches from one niff to another to make mash ups on weapons. Scales and transformations on XYZ are saved
- export to blender, by following hexabits instruction, should work. I'm gonna try again this evening and let you know.
I am having issue with a large portion of the file "Starfield - Meshes01.ba2" (i have tried the laser turret, starborn helmets, and Armillary) when i import them the scene resets but doesn't show the model, however when i do click "0.Ninode" it dose show the models rig, both for models from the archive and BAE.
I am using 2.0 dev9 that is within the "NifSkope_2_0_2023-09-29-x64" file.
I do not know if it effects anything, but I have both nifskope and starfield on a external SSD.
Did you add
Starfield - FaceMeshes
Starfield - Meshes01
Starfield - Meshes02
Starfield - MeshesPatch
BA2s to your Resources > Archives tab?
Also, I recommend putting MeshesPatch
at the top of the archive list, so it's loaded from first.
Did you add
Starfield - FaceMeshes Starfield - Meshes01 Starfield - Meshes02 Starfield - MeshesPatch
BA2s to your Resources > Archives tab?
Also, I recommend putting
MeshesPatch
at the top of the archive list, so it's loaded from first.
I didn't, and when i did it worked! thank you =D
Thank you so much for your work on this!
TLDR: 1. Are you aware of a reliable NifScope -> 3dsMax->nif pipeline? 2. would the FBX SDK cause more problems than it would solve?
...
I've been trying to get a 3ds Max workflow ironed out. I can import from the ba2's, export to glTF, import into Blender, Export as FBX, import into Max, export as .Nif. Upon reloading the exported nif into nifscope the skeleton is disassociated with the mesh and the reference export/root nodes are missing... the geometry (all 3 LOD) seem to re-import fine, aside from the scale, though I think that is a blender export setting.
Are you aware of a better pipeline? (I'm using 2021 - no glTF i/o) Have you looked into the Autodesk fbx SDK? It's supported by 3ds Max, Maya, Blender, and Lightwave. The docs claim it is able to handle the nodes that gltF doesn't natively support, also it claims a built in collada and OBJ i/o component. (VS2017-2022)https://www.autodesk.com/developer-network/platform-technologies/fbx-sdk-2020-3-4 or (VS2012-2017)https://www.autodesk.com/developer-network/platform-technologies/fbx-sdk-2020-0
I'm willing to be a test subject for 3ds Max i/o if need be.
Hi there. ATM, there is a working blender plugin on the nexus https://www.nexusmods.com/starfield/mods/4360
Using this in combo with nifskope should work.
I'm gonna test tonight and tomorrow again, I'll tell the workflow I'm using.
0) Open niff in nifskope, see which geometry I want to edit 1) Open the geometry in Blender, edit, export (in the tool menu, not from "File" 2) Rename the geometry, so you do not overwrite anything. I usually change the latest two hex number, increasing 3) Change the geometry in the nifskope line, writing the new file name, and save it 4) Close nifskope, copy & paste folder in starfield installation data\geometry[correct path] (nifskope do not see documents) 5) Open again the nif in nifskope, now you should see the edited geometry loaded 6) DO NOT MIX DIFFERENT PART FROM DIFFERENT NIFs IN ONE GEOMETRY, if you want to do mashups, do it in nifskope copying branches, or it will fuck up all body weights.
Let me explain, you you want to make a new nif with marine spacesuit arms and security armor chest, first you edit the arm, you can import them together so you can see clipping to fix, etc, but when you export, export two different geometries, load and save two different nifs, then copy one branch into another nif and all will be fine. If you export arms and body in the geometry of the body, chest body weights will be fine, while arms body weight will be all wrong.
7) Now that you have exported your mashup, remove the geometry folder from installation folder, it appears it crashes the game, at least to me 8) open the game, you should be able to see your new mesh.
I've already made my first new mashup last night, still have to publish it because I want to do a minor fix, but this workflow should be fine.
Could it be that there are still parts (values?) NIFskope cant identify and therefore they dont showup under the 0Ninode or child list?
I'm trying to figure out ship snappoints but cant find anything in the NIF that points to snappoints!?
What I did find out, I flipped a ship module in NIFskope and replaced an existing shipmodule with this flipped NIF.
What happened was it showed up in game like i wanted but inherited the snappoints of the NIF I replaced. So the NIF i edited originally had snappoints aft and top, after replacing it had the snappoints fore and aft like the module i replaced.
This does make me question if the snappoints are in the NIF like they are in FO4, I watched Kingaths tutorial on FO4 snappoints and howto add them using NIFskope to see if that gave me a lead!?
I did find a number of markers related to snappoints \meshes\markers\shipmarkers: shipsnapnodemarker_bottom01 / shipsnapbehaviorhelper01 etc.
Again though no link between these and the NIF of the modules!?
I'm new to this so I am learning as I go.
Thanks in advance
P.s. using NIFskope 2.0 Dev 9
I'm trying to figure out ship snappoints but cant find anything in the NIF that points to snappoints!?
Because they're no longer in the NIFs, they are in the ESM.
I'm trying to figure out ship snappoints but cant find anything in the NIF that points to snappoints!?
Because they're no longer in the NIFs, they are in the ESM.
And we cant do much with the ESM yet :-(
I'll get the ESM explorer I saw on Nexus tomorrow, see if I can at least locate them
:-) Thank you very much for this info
Sorry for bumping this thread, but I have no idea where else to ask. Is there any plans to support seeing textures in NifSkope for Starfield? I use NifSkope for finding nice models, do a quick preview, do placeholders for snaps, but I really miss seeing textures. Some models look great in theory, but once I see them in game they don't fit my idea, so waste of 30 minutes. If anyone has a timeline, workaround, or any pointers at all please share. If it involves exporting to Blender that's fine, it's still quicker than adding to game.
For now, there is a fork here that can display Starfield textures with some limitations (only the first layer of the material is rendered), and it makes a few fixes/improvements to Fallout 76 rendering as well. It also shows detailed material properties like texture paths, although editing them is not possible, only material paths (BSLightingShaderProperty block names) can be changed.
For now, there is a fork here that can display Starfield textures with some limitations (only the first layer of the material is rendered), and it makes a few fixes/improvements to Fallout 76 rendering as well. It also shows detailed material properties like texture paths, although editing them is not possible, only material paths (BSLightingShaderProperty block names) can be changed.
This sounds as enough for my needs. But... is it possible to provide a ZIP with binaries? It had been years since I compiled anything, and all I see is source. Would be most grateful! Doesn't have to be proper github release, any downloadable ZIP is fine
This sounds as enough for my needs. But... is it possible to provide a ZIP with binaries?
Binary packages are generated automatically by GitHub on any update to the source code. If you open the most recent successful workflow run under actions, they are available as "artifacts" (build-linux and build-win) in .zip format. Downloading those files requires signing in to GitHub.
This sounds as enough for my needs. But... is it possible to provide a ZIP with binaries?
Binary packages are generated automatically by GitHub on any update to the source code. If you open the most recent successful workflow run under actions, they are available as "artifacts" (build-linux and build-win) in .zip format. Downloading those files requires signing in to GitHub.
Ah ok, found them, didn't know that even existed. Anyway, I get error about shaders on start. I open NIF and can see eg connection points but no model displayed (I can see data though).I've tried disabling shaders, no error in that csae on startup, but still don't see the model either. Seems it pulled all settings from dev version posted here earlier, one I've been using so far. Sorry if this isn't helpful :-/ If you point me in the right direction I can try to debug it and get you logs or whatever. W11 Pro btw.
Anyway, I get error about shaders on start.
Can you post the shader compilation error message?
Anyway, I get error about shaders on start.
Can you post the shader compilation error message?
Sure:
There were errors during shader compilation
stf_default.frag:
ERROR: 0:446: 'variable indexing sampler array' : not supported with this profile: none
ERROR: 0:446: '' : compilation terminated
ERROR: 2 compilation errors. No code generated.
stf_default.prog:
depends on shader stf_default.frag which was not compiled successful
I was adding "Connection points" in the dev9, and that renders fine (but without textures) on the version from post above.
In your version it looks like this (after dismissing shader error):
Does it fix the error if you replace this line at the beginning of stf_default.frag:
#version 130
with one of these, either:
#version 150 compatibility
or:
#version 400 compatibility
Does it fix the error if you replace this line at the beginning of stf_default.frag:
#version 130
with one of these, either:#version 150 compatibility
or:#version 400 compatibility
Yes, both get rid of the shader error. But I still don't see the mesh or anything in the preview window, only little green dots where the connection points should be.
Does the mesh disappear if you use the modified stf_default.frag file with the original dev9 build of NifSkope (or vice versa, the old dev9 shader with my fork)?
When I move your file to original dev9 build (I moved only stf_default.frag
), then model is rendered in dark grey, while usually it's almost white.
But when I move original dev9 version of stf_default.frag
to your build, I still don't see the model.
Can I do anything else? And does the change of color indicate anything?
I can reproduce the issue by running on an AMD GPU (instead of Nvidia) with the open source drivers. In fact, on Linux, even the original dev9 version freezes now on trying to render anything, but the Windows build works with Wine and has the same problems as you reported (it does display Fallout 76 models, however). Now that I can reproduce the issue, I will look into fixing it, and ideally the Linux AMD driver freeze as well that is not specific to my fork. Edit: the latter was fixed by reducing anti-aliasing, so it was not necessarily a NifSkope bug.
There is an updated build available now that tries to work around the issue.
I'm making this issue primarily as a PSA to anyone hoping for or expecting quick turnaround time for Starfield support. TL;DR - I can quickly get files loading in NifSkope, but nothing would be visible in the viewport.
Starfield introduces a number of issues for existing tooling. The actual geometry is now external to the NIF. This happens in at least two ways:
BSWeakReferenceNode
A list of many NIF references via their BSResourceID (file hash, extension, dir hash), plus transforms to place them into the root scene. This can be dozens or hundreds of separate NIFs loaded in one file. This is similar to
BSDistantObjectInstancedNode
introduced in FO76.BSGeometry
Holds no geometry itself, includes a 41 character filepath (some kind of hash) to a .mesh file which stores the geometry. BSGeometry can reference more than one .mesh (up to 4).
NifSkope and nifxml can be updated for the format changes. However, there will be no visible geometry in virtually all of the NIFs without a refactor to NifSkope's serialization, rendering, and data management.
Making this block functional (i.e. visible) in NifSkope may take an overhaul to the rendering and scenegraph. The solution I foresee is adding an actual Scenegraph pane. This will hold many loaded NIFs, and selecting a NIF will make its data active in the Block List. The NifModel will almost certainly need to be unloaded because of how much memory it takes up so that there is still only one NifModel at a time. This means the geometry will have to be loaded and cached for non-active NIFs.
Such a refactor will also be beneficial to
BSDistantObjectInstancedNode
and things such as loading more than one BTO/BTR file (e.g. loading an entire map's terrain).The null-terminated array is awful but that is how they implemented it. It is thankfully encapsulable in nifxml.
For rendering, another shape class will need to be introduced, since the others are heavily coupled with NifModel. Though a workaround in lieu of visible meshes would be to provide actions to export/import the .mesh to/from another format for editing in Blender etc.
Other bookkeeping:
Dummy bones (flat root bones) in NIFs seem to have been replaced by:
Adjusting bones via dummy bones has been replaced by: