• Which the release of FS2020 we see an explosition of activity on the forun and of course we are very happy to see this. But having all questions about FS2020 in one forum becomes a bit messy. So therefore we would like to ask you all to use the following guidelines when posting your questions:

    • Tag FS2020 specific questions with the MSFS2020 tag.
    • Questions about making 3D assets can be posted in the 3D asset design forum. Either post them in the subforum of the modelling tool you use or in the general forum if they are general.
    • Questions about aircraft design can be posted in the Aircraft design forum
    • Questions about airport design can be posted in the FS2020 airport design forum. Once airport development tools have been updated for FS2020 you can post tool speciifc questions in the subforums of those tools as well of course.
    • Questions about terrain design can be posted in the FS2020 terrain design forum.
    • Questions about SimConnect can be posted in the SimConnect forum.

    Any other question that is not specific to an aspect of development or tool can be posted in the General chat forum.

    By following these guidelines we make sure that the forums remain easy to read for everybody and also that the right people can find your post to answer it.

my textures are huge

Messages
112
Country
unitedkingdom
i am making a scenery object for the merced river, about 3km x 1km. i'm using 7 4kx4k tiles on a single model. each tile, although quite large, 16MP, is mostly nothing. i started with .png file but the files were quite large, about 12MB each, after .dds conversion, 21MB.

so i converted the .png to .jpg and now they are about 1.5MB each. but when i convert to .dds, they are 10MB each.

given that about 90% of each image is nothing, is there a better way to make the .dds?

BTW, FSX doesn't seem to be having a problem with it, no change in frame rate at all

BTW, i know why jpeg compresses nothing so well, it uses DCTs so low frequency produces very few bits. i assume DTX doesn't, that's why i'm asking, if there is a better method of dealing with large textures with large blank areas.
 
Last edited:
Hi Steve:

I'm not certain what aspect of your scenery project may require ground and water bodies to be 3D models, but typically a valley floor may be implemented with better impact on performance (even if using higher resolution) via FSX SDK "Terrain" methods:

* custom terrain mesh (higher-resolution than local 10-meter mesh from OrbX freeware / payware)

* custom photo-real land class textures (mesh-clinging; higher-resolution than local 1-meter mesh from OrbX freeware / payware)

...via source data submitted in the required format for processing by FSX SDK Resample.


If you wish to change the Merced River 'shape', you can instead do this by means of "flatten" polygons which modify terrain mesh display

...via source data submitted in the required format for processing by FSX SDK SHP2VEC.


Regarding the higher-resolution (and therefore 'larger') textures used for a proposed 3D 'ground' object, these can instead be applied to the underlying ground surface provided by the FSX terrain mesh (...either a default or 3rd party / custom BGL file).


AFAIK, when a blend mask is used in one of the (2) available Alpha channel 'slots' of a GeoTiff / TIFF file format:

1 - black (0,0,0) 8-bit TIF Water Mask enables FSX "water" (with wave movement, splashing and/or sinking of user aircraft etc.)

1 - gray (ex: less than 255,255,255) 8-bit TIF Blend Mask allows the photo-real texture to show through the FSX water from underneath

https://msdn.microsoft.com/en-us/library/cc707102.aspx

https://msdn.microsoft.com/en-us/library/cc707102.aspx#Example4CustomTerrain


An example INF file using GeoTIFF source imagery for custom photo-real land class textures < thanks, Dick ! ;) > :

http://www.fsdeveloper.com/forum/threads/seasonal-photoreal.3783/


CAVEAT#1: IMHO, it is better to preserve graphical image visual quality by converting PNG to BMP / DDS rather than using JPG as an intermediate format, as it is typically very "lossy".

CAVEAT#2: Always back up source imagery files, and work with COPIES

CAVEAT#3: Always back up Geo-referencing info associated with GeoTiff files BEFORE opening them to perform edits in a graphics application, then restore Geo-referencing info to GeoTiff files when finished ...via a special GIS utility.

GeoTiff files can be used as a "1-piece" work file for many aspects of FSX SDK terrain scenery work which reduces risk for errors (and workload) when creating source data and INF files etc. for FSX SDK Resample


NOTE: To eliminate un-necessary raster info (from a output BGL) for colors which will not be 'visually' displayed when utilized as part of the source imagery for ex: custom photo-real land class textures (mesh-clinging), a parameter is utilized in the INF file submitted to FSX SDK Resample along with the imagery which will be 'visually' displayed:

NullValue=xxx,xxx,xxx

http://www.fsdeveloper.com/forum/threads/changing-null-value-in-sbuilderx-inf-file.427714/

http://www.fsdeveloper.com/forum/threads/sbuilderx-water-mask-issue.426022/


Another variation of the "NullValue" parameter may be used to recursively process raster data in a multi-layer / multi-source FSX SDK Resample compilation scenario:

NullValue=,,,,0

I would advise leaving the water colour in your source image. The white border is bleeding into the final result as a result of the white in the photosource.

Similarly, use a non-RGB Null value, or omit the line entirely.

NullValue = ,,,,0

for example, will omit data points filtered out in the blend channel.

http://www.fsdeveloper.com/forum/threads/my-first-photoscenery.427289/
LandWater masks are only single bit, when resampled, so the edge can be dithered, but greyscale won't work.

Often a light 'feather' to the edge of (say) 2px provides enough softness to smooth the line.

Blend boundaries (as a rule):

On land, follow edges of features on the picture (never the photo edge!) and use a very small feather margin (eg 2px)

On water a 'soft' greyscale gradient or feather is best to gradually mix the underlying water.

Also:

With your imagery, it needs to increase sat, and reduce brightness. Alternatively, add 'selective' black in a colour layer.



BTW: An excellent tutorial on using SBuilderX to create photo-real scenery which discusses many incidental tips /tricks:

http://www.flightsim.com/vbfs/showthread.php?250762-How-to-create-photoreal-scenery-for-FSX


FYI: Some related info on BMP / DDS files

http://www.fsdeveloper.com/wiki/index.php?title=Texture_formats_overview

https://msdn.microsoft.com/en-us/library/windows/desktop/ms536393(v=vs.85).aspx


Hope this info helps to sort out your options, and to 'minimize' any resulting BGL output file sizes. :)

GaryGB
 
Last edited:
hi gary, i did watch that video on sbuilderx but the issue with the river is in the orbx model the river valley is indented, which in real life, it is of course, its a river, but the water texture sticks to the land surface so it looks bad up close, plus the resolution isn't that great.

i was under the impression from one of your other replies that i can't edit the base terrain DEM data so i made an object in blender. plus, i want to get the bridges in, especially sentinel and stoneman bridges, but in real life, they are too small to fly a plane through so i'm going to have to use a bit of artistic license to make them edge 540 size.

its late here now but i'll read your post in detail tomorrow

thanks again
 
hi gary, i did watch that video on sbuilderx but the issue with the river is in the orbx model the river valley is indented, which in real life, it is of course, its a river, but the water texture sticks to the land surface so it looks bad up close, plus the resolution isn't that great.

i was under the impression from one of your other replies that i can't edit the base terrain DEM data so i made an object in blender. plus, i want to get the bridges in, especially sentinel and stoneman bridges, but in real life, they are too small to fly a plane through so i'm going to have to use a bit of artistic license to make them edge 540 size.

its late here now but i'll read your post in detail tomorrow

thanks again

As alluded to above, you may discover when you study and test the use of ex: SBuilderX, that one can create:

* "flatten" polygons (compiled to a CVX-type BGL via FSX SDK SHP2VEC) as a vector layer

...super-imposed in the SBuilderX workspace on a background map which displays:

* aerial imagery (also compiled to a mesh-clinging custom land class-type ground texture / placement BGL via FSX SDK Resample)


FYI: The aerial imagery can of course be higher-resolution than the local 1-meter custom photo-real land class textures from OrbX freeware / payware) if available for download from GIS data servers and/or other tile servers.

If such higher-resolution aerial imagery is not available for download, one can 'create' it as "synthetic textures" (ex: graphically-derived from a mix of other aerial imagery sources).


"Flatten" polygons can be used to define the elevation and geographic placement of all individual coordinates of vertices for the edges of the river bed, as well as the elevation of the water surface within the river bed, as the single 'flatten' polygon will "cut into" any local underlying terrain mesh to the depth that you define as a relative Altitude / elevation value, by specifying that value as Above Ground Level aka "AGL" (...and not relative to Mean Sea Level aka "MSL" or Above Sea Level aka "ASL").

Thus, rather than needing to "edit the base terrain DEM data", one simply loads a local 'flatten' BGL which defines terrain shape by modifying at run time in FSX, terrain mesh display for ex: the Merced River bed from underneath the super-imposed / mesh-clinging higher-resolution aerial imagery custom land class-type ground texture BGL. :idea:



BTW: A bit of artistic license can add to the "fun" in ex: a race mission; dare I say missions can be /should be ..."entertaining" ? :duck:

NOTE to P3D users: "Academically-challenging to the aerobatic maneuver learning process" can be substituted ...for "fun" ! :rotfl:

GaryGB
 
Last edited:
so i looks like i need to learn yet another tool! i already patched together a bunch of hi res images of the merced river using photoshop's cool automated 'photomerge' feature, tiled them into 4kx4k blocks and mapped them onto 3d meshes

since i got the 3d model method working already and i'm keen to get back to race circuit building, i think i've leave SBuilderX for later, i will take a look however

(BTW, i figured i can put more than one river segment into a 4kx4k image. so each image is still about 11MB but i need less textures.)
 
Last edited:
hi again Gary, you are absolutely right of course, modeling a large, mostly flat area in blender and trying to place it is a really bad idea. i need to throw that away and start over with SBuilderX

i keep looking at the amazing deepzoom site you sent me a link to a few posts ago http://www.xrez.com/yose_proj/yose_deepzoom/index.html

i would so like to redo the valley from scratch with this resolution but i know that would be weeks, possibly months of work. it would be pretty amazing if i every got it however. i understand SBuilderX can handle the flats to medium slopes just fine, even at high texture resolution right? but i would still need to make 3D objects like i'm doing now for faces above about 70-80 degree slope right since the base terrain system is intrinsically a top down system and doesn't handle near verticals well and true verticals, overhangs and tunnels at all. is that basically correct?
 
Hi Steve:

The "stretched rubber sheet" distortion when aerial imagery is draped "top-down" onto the terrain mesh as textures via the FS world rendering engine is indeed a major factor which might compel use of 3D models for valley walls that have a near vertical orientation relative to the "ground" of the valley floor.

Another factor is that the real world camera viewpoint used when the above panoramic landscape photos were taken was via a "horizontally oriented" perspective parallel to the plane of the valley floor (thus perpendicular to the vertical plane).

Therefore those ultra high detail images would IMHO need to be draped via a "lateral approach" to avoid the same undesirable texture distortion in FS at run time that one would have incurred when draping aerial imagery "top down" onto the vertically oriented (sky-ward) surfaces of 'terrain' objects ...whether displayed by means of terrain mesh, or by 3D objects.

However, IIUC, you plan to utilize ultra high resolution textures only on the major landmarks with predominantly vertical faces for which suitably detailed "horizontal landscape" imagery is available, so that might lighten the workload by allowing the remainder of the valley terrain structures and ground surfaces to be implemented by aerial imagery draped "top down" onto a custom higher resolution terrain mesh as custom photo-real land class textures, such as is typically done in the FS world rendering engine.


Creation of custom terrain mesh and photo-real land class textures can be a mostly automated process assuming availability of suitably detailed, georeferenced, orthographically projected etc. data from sources which are free or affordable, and which allow unrestricted use of the data for ones intended purpose.


One might wish to initially determine what approximate target resolution the available imagery is to be compiled into for use within FS by analyzing the size in Meters of the horizontal span of the landscape / object(s) versus the horizontal span in Pixels of the imagery, whether aerial "top down" or "horizontal" panoramic ...to determine the effective LOD of textures to be draped onto terrain mesh (or mapped onto 3D objects), and how many tiles of ex: 4096 x 4096 Pixel imagery are needed to cover the areas / objects of interest.

In the case of source imagery for custom photo-real land class textures, one will end up deriving the "size" (aka 'dimensions') of imagery Pixels (expressed as Pixels per Geographic 'Arc Degree') for use in the INF file submitted to FSX Resample.

[EDITED]

https://msdn.microsoft.com/en-us/library/cc707102.aspx#SourceParameters


CAVEAT: FSX SDK docs infer that achieving 'intended' run time display of higher resolution textures may require use of resolutions no greater than LOD-13 (QMID-15) aka 4.75 Meters/Pixel:

https://msdn.microsoft.com/en-us/library/cc707102.aspx#QMIDandLODValues


The above FSX SDK doc caveat may be particularly true if the user aircraft is moving at higher rates of speed (such as in a "race") at a close distance to the textured object.

Thus, when a FS user aircraft is moving very fast in close proximity to the ground or a 3D surface, rendering of very high resolution textures may be at risk for "blurries", slowed display and/or clipping / shifting of LODs to lower resolution MIPMAPs etc. if they are mapped over a very high vertex density 3D model, or draped over a very high grid density (higher LOD / smaller distances between elevation data points) terrain mesh.

This may be even more likely to occur if the end user computer running MSFS is not powerful or well-optimized. :pushpin:

However, the above FSX SDK doc caveat is rather curious, as LOD-13 is the maximum resolution that FS2004 (aka "FS9") can display for custom photo-real land class textures, and the FSX default land class terrain texture resolution is 1.2 Meters/pixel (LOD-15); why would ACES have given us high resolution ground textures with a rendering engine that "can't keep up" ? :scratchch

Regardless, over the years that have passed since FSX was still in development prior to its release, the FS Community has shown that we are quite capable of using and displaying resolutions higher than the supplied "out-of-the-box" max 30 Meter / LOD-10 default terrain mesh and 1.2 Meter / LOD-15 default land class terrain textures ...on "most" modern computers running FSX, without incurring the above rendering and display issues (as ACES likely had already envisioned). :idea:

[END_EDIT]


NOTE: SBuilderX can perform most of the above required work and calculations semi-automatically.


Furthermore, by using GeoTiff format for source data submitted to FSX Resample to make terrain mesh, ones workload with such calculations can be greatly reduced or eliminated.


I believe you may find that some work with SBuilderX may prove to be worth the investment of time, such that you might find other aspects of your perceived pending workload can be reduced substantially. ;)


And as for draping texture materials onto Yosemite landmark 3D objects, IMHO, the method I previously described using Sketchup (with optional use of 1 or more Ruby script plugins) may merit further consideration.

http://www.fsdeveloper.com/forum/threads/mesh-from-sketchup.433127/#post-701342


This procedure might make it possible to complete the texture material draping operation onto the side wall structures of those Yosemite landmark 3D objects via a "lateral approach" in a single step (and/or a very few additional such steps), as shown in Taff Goch's tutorial. :)

https://3dwarehouse.sketchup.com/model.html?id=7b8a0bd3d848e985e523d8740930d7f9

http://help.sketchup.com/en/article/94883

GaryGB
 
Last edited:
hi gary, so i finally put El Capitan to bed after about 8 rebuilds. its not geological time but it started to feel like it. anyhow, i did a cool trick where i added a very dense 4k texture area where you can see individual climbers, and put that directly under the race gate where your speed tops out, so its looking very nice now, all the way up the face, even down to a distance of about 5 meters. i'll make a race circuit video soon.

i downloaded SBuilderX and after figuring out i also needed the google dll, got it working
so i've been looking at this http://www.flightsim.com/vbfs/showthread.php?250762-How-to-create-photoreal-scenery-for-FSX
it looks like a lot of work to make blend marks and put polys to level out the river water surface. i think for now i'm just going to add a few bridges and be done with it. i can rework the river later.

steve
 
Back
Top