I have two versions of the same object, with identical geometries but differing UV sets.  Both UV sets have the same islands, but these are rearranged and rotated.  I have an existing texture for the first UV set, but need one for the second.

Does anyone know of a tool that can convert the texture from the first object to the second?  I seem to have lost my google-fu, since they modified all the search engine algorithms to use AI to second guess the user.  I've been largely inactive in Poserdom for long enough that I honestly do not know where I might ask such a thing, if not here.  Which is to say, I hope this is not an off-topic question, and I apologize if it is.  (I would formerly have placed a sad & embarrassed smiley here, but I've lost track of how those are used nowadays.)

If I remember correctly, UVMapper can save out the uv map from one (.uvs) and then read it in and apply to the other obj.

Bill

Blender can do it. I can dig out a YouTube tutorial I saw recently if you think that’s an option for you.

willyb53 posted at 8:24 PM Sat, 17 December 2022 - #4451987

If I remember correctly, UVMapper can save out the uv map from one (.uvs) and then read it in and apply to the other obj.

Bill

You would then import the target obj and apply the saved out uvs file to the target obj. I'm not sure if you need to delete the pre existing UVs from the target or if applying the uvs file will overwrite. 

Which makes me wonder if it's possible to simply copy the vt lines block from the source uv obj and paste into the target obj file (with the original vt block taken out of course) via a text editor... obj files are dead simple to edit.

Cross post with odf, apologies...

Thanks, everyone.  I may not have been clear about my problem.  I don't need to transfer UV sets between objects, I need to convert a texture graphic designed for one UV set so it can be used with the other UV set.  I need to move the pixels around on the texture map image (not a UV template, but an actual texture), so they fit into the new spots on the modified UVs, right.

Ten years ago I would have spent a month fighting with PoserPython and PIL and cursing a lot, to try to build such a thing.  My programming skills, like my google-fu and my general internet communication skills, are pretty rusty these days.  Dang.

I reiterate, Blender can do what you want. :-)

Cage posted at 10:01 PM Sat, 17 December 2022 - #4451992

Thanks, everyone.  I may not have been clear about my problem.  I don't need to transfer UV sets between objects, I need to convert a texture graphic designed for one UV set so it can be used with the other UV set.  I need to move the pixels around on the texture map image (not a UV template, but an actual texture), so they fit into the new spots on the modified UVs, right.

Ten years ago I would have spent a month fighting with PoserPython and PIL and cursing a lot, to try to build such a thing.  My programming skills, like my google-fu and my general internet communication skills, are pretty rusty these days.  

I know how I would do it.

odf posted at 10:12 PM Sat, 17 December 2022 - #4451993

I reiterate, Blender can do what you want. :-)

Can you tell me anything more?  My version of Blender is from... maybe 2008.  My computer, for that matter, is from 2011, always offline and running Windows 7.  My version of Poser is PPro14.  Cage is super-way out of date.  Which makes me wonder whether I can run the latest versions of Blender.

Well, here's the tutorial: https://www.youtube.com/watch?v=wsO1eozb1Qk

My previous computer was from 2014 and couldn't run recent versions of Blender. Then again, it was running Linux, which tends to aggravate driver issues. Myself, I haven't used Blender in ages before getting that new computer two years ago, so I have no idea which version one would need.

At any rate, the method relies on two things: one, Blender can "bake" a node-based texture into an image map, and two, it can use different UVs in different nodes of the same shader, so it can use on set of UVs to generate a texture and another one to bake it to.

The part that's missing from the tutorial is how to load to different UVs from files and apply them to the same geometry (in the tutorial, one is generated from scratch right within Blender). In theory, that should be easy and obvious, but with Blender, things rarely are.

Cage posted at 10:01 PM Sat, 17 December 2022 - #4451992

... I need to convert a texture graphic designed for one UV set so it can be used with the other UV set.  I need to move the pixels around on the texture map image (not a UV template, but an actual texture), so they fit into the new spots on the modified UVs, right.

An all Poser solution is to apply your source image to the ambient or (better) altDiffuse channel of the texture of a 'cloth plane' and apply magnets, masks and morph brush to rearrange the uv's as you need them when viewed by a top camera with long focal distance. Then render to get the new map.

It comes set up for that specific task, with some specific bells and whistles to match tone etc. so it looks more complicated than it is.

It takes some time to set up, but once it is set up for a uv conversion like V4->PE it can do any source image swapped on the source pane. 

FVerbaas posted at 10:11 AM Sun, 18 December 2022 - #4452023

Cage posted at 10:01 PM Sat, 17 December 2022 - #4451992

... I need to convert a texture graphic designed for one UV set so it can be used with the other UV set.  I need to move the pixels around on the texture map image (not a UV template, but an actual texture), so they fit into the new spots on the modified UVs, right.

An all Poser solution is to apply your source image to the ambient or (better) altDiffuse channel of the texture of a 'cloth plane' and apply magnets, masks and morph brush to rearrange the uv's as you need them when viewed by a top camera with long focal distance. Then render to get the new map.

In my freebies here there is an all-poser solution that applies the technique to morph V4 face textures to suit Project Evolution: face-texture-converter

It comes set up for that specific task, with some specific bells and whistles to match tone etc. so it looks more complicated than it is.

It takes some time to set up, but once it is set up for a uv conversion like V4->PE it can do any source image swapped on the source pane. 

What does it mean when one sees "Heads Up! This user's messages are blocked from your view." in a thread?  Are we blocking one another now?  Seems like a way to create discussions full of holes & gaps.  The internet has gone all pear-shaped, since the advent of these wander-phones.

That's been happening with primorge with several others including me.  Something must be up with his profile because he's the only one that has been happening to.

You could have blocked the user by mistake.  Normally if you find a user that you don't want to read what they post you press the little speaker icon next to the yellow info icon.  then you will see that you have this user ignored.  If you see the Heads up you can click on the unignore this user and you will then see their messages.  

Ah, so I need to be careful where I click, as I scroll a page.  That would explain the matter.  I often fidget-click while reading, not paying attention to where the cursor is when I click.  And I should not do that, I guess.

Here's another option: a stand-alone Windows utility which is fairly ancient and should work on any machine.

http://www.uv2uv.spacymen.com/

It does exactly what you want, assuming I've understood you properly. The inputs are two OBJ files composed only of triangles, with your source and destination UV mappings; and a texture map for the source OBJ which has to be in BMP format. The output is the map transformed to fit the destination OBJ, also in BMP format.

Converting back and forth is an extra few steps, but easy enough to do. UVMapper will triangulate the input meshes for you; you don't have to keep those interim files of course.

Woohoo!  Thank you, EnglishBob!  This one seems straightforward enough for me to try it and report back later today.

The uv2uv program does exactly what I need, and this user found its website delightful, with the frame and the program logo that looks like it might have been made in Amorphium or something.  Really lovely.  This saved me, potentially, hours of repainting hair textures.  All I had to do was pull the resulting textures into Modo, so I could expand the UV seams.

Glad to have been of service! I tend to prefer small, simple utilities that do one thing - that said, I'm intrigued to hear that Blender can do it. Is there anything that Blender can't do? (Not trying to start a challenge, by the way.)

Off-topic, you might like to consider editing your signature block at some stage. Not that I mind being thanked for things, but it's vaguely embarrassing when I stopped doing those things three years ago. :) This is where you go to do that: https://www.renderosity.com/settings/forums

It fails a more challenging test rather drastically.  An attempt to convert an Antonia Standard texture to the A-mapping made a real mess.  I suspect the different grouping of the two Antonia obj files throws off the polygon, tri, and UV indexing.  I'll try prepping the Antonia files, for a better test.

EB, your avatar image reminds me of the little white dot that would appear, as the old CRT televsions were shutting down.  As Neil says, "It must be a really old telly."

It fails a more challenging test rather drastically.  An attempt to convert an Antonia Standard texture to the A-mapping made a real mess.  I suspect the different grouping of the two Antonia obj files throws off the polygon, tri, and UV indexing.  I'll try prepping the Antonia files, for a better test.

Agreed, you might try saving your triangulated interim OBJ with only one group. I've only ever used uv2uv on a single-group mesh, and it was hair, coincidentally.

EB, your avatar image reminds me of the little white dot that would appear, as the old CRT televsions were shutting down.  As Neil says, "It must be a really old telly."

Not that I've tried anything in practice yet, but I think overlapping UV islands will be problematic for all of the methods mentioned so far. Probably best to convert the target UV set to UDIMs first and then, if the tool used does not support UDIMs directly, convert the texture one tile at a time with the necessary adjustments done based on the UDIM version of the UVs. I hope that makes at least a little bit of sense.

At any rate, since Antonia 1.3 no longer supports the old UV mapping, I've been considering converting some of the textures available for Antonia-A to the standard UV sets. Not sure yet which method I'll try (Blender, an automated variant of FVerbaas' Poser-only method, or an external tool), but if I get anything working, I'll report back here.

Cage posted at 2:42 PM Tue, 20 December 2022 - #4452246

The uv2uv program does exactly what I need, and this user found its website delightful, with the frame and the program logo that looks like it might have been made in Amorphium or something.  Really lovely.  This saved me, potentially, hours of repainting hair textures.  All I had to do was pull the resulting textures into Modo, so I could expand the UV seams.

Working with Bake Items and Baking Tools (modo)

Ooh.  Thanks, DCArt.  I have Modo 302, purchased during a short-lived period when I actually have money to spend on such things.  I will have to see if my version has the tools to work with your link.

May not go back that far. I think that's the first version I got and I can't remember it being there.

Edit: Looks like it was introduced in Modo 10

I just learned that apparently DAZ Studio has a built-in (?) tool for this, see video here: https://www.youtube.com/watch?v=YxE5Lfkid1M.

I haven't tried it or watched the video with full attention, so I'm not sure what the limitations are. But they do address overlapping UV islands by letting the user assign material zones to "templates" and then baking one image per template.

Update: I tried the DAZ tool on Antonia using my somewhat dated version of Studio 4. Unfortunately, the results came out all scrambled. Since I'm not a DAZ user, I won't be digging any deeper.

Over the holidays I've made some progress automating the Poser-based solution FVerbaas described. I can't call it Poser-only anymore because I use a stand-alone Python script to create a transfer object from the two versions of the original geometry. Load that into Poser, load the textures, do a bit of extra fiddling (some of which I can also delegate to the script to make things easier) and render the different texture tiles one by one. The image shows one of BlueEcho's textures for the AntoniaA UV sets converted to the standard UVs.

Looks good! 

Any plans to do same for the V3 mapped version? That would bring Tony a plethora of skins. Some may have baked highlights as customary in the day, but still useful.

Looks good! 

Any plans to do same for the V3 mapped version? That would bring Tony a plethora of skins. Some may have baked highlights as customary in the day, but still useful.

Ah, I misremembered. Turns out I have a handful of V3 textures to choose from. I still might have a look around for something more unusual, so that I'm more motivated to do the testing.

As a quick test, here is Antonia 1.3 wearing the Sophie texture for V3 by 3dStrike. Some visible seams, as expected. I imagine the quickest and dirtiest fix would be to use a background that approximates the skin color when rendering out the transferred textures. Anyway, I think we have proof of concept. Also, I'll be happy to share that Python script once I've made some improvements to it.

If this is rendered in Poser, did you set the skinning to unimesh?

If this is rendered in Poser, did you set the skinning to unimesh?

Here's another test render, this time with Morris' HyperReal texture, showing more clearly that the issues are at the texture seams, not the body part seams.

I think it would be relatively easy to postprocess the textures after rendering them out to produce some "bleeding" of the texture islands into the surrounding empty texture space, and there are probably tools that can do it automatically. But that's an extra step for the user, which is a bit annoying. So ideally my script would create the transfer object in such a way that the bleeding is produced automatically when rendering out the converted textures. That's a little bit more involved than the postprocessing but should still be doable. Mostly a question of whether I can be bothered...

Anyway, I've made the other improvement, which was to "UDIMify" the texture tiles so that when the transfer object is loaded into Poser, they're nicely separated as shown below. Much easier to work with.

What if you process a V3 uv template and compare the result with Antonia's? That should give you an idea of how to solve.

When there is a shift of the uv islands to one side there may be a rounding off error.  

When the uv islands are a tad too small some logic is needed to grow the uv islands a bit.

Allow me to give some idea for this: A seam means a 3D vertex is represented by more than one vertex in 2D, so that's a way to identify uv island edges.

Orientation of vertices around a facet follows right-hand rule so in the morph target an edge leading from one edge vertex of a facet to another needs to be shifted perpendicular to that edge, negative normal direction. To avoid doubles this shift should not be done for the last edge vertex of the facet because that one will have it's shift with the adjacent facet.  

Addition: Growing the target uv islands, or shrinking of the source islands of course can be implemented as morphs in the object.  

What you're seeing is a problem I've had with Modo and texture painting, for years.  Modo has a built-in feature which extends the pixel-colors at the UV seam edges by an unspecified increment, which has always solved it for me.  As reported above, I had the same issue and solution for uv2uv.  I assume it's some sort of float precision issue, with the various UV methods being more precise than Poser prefers about finding the edges of the UV islands, but I can only guess.

The solution to getting rid of the seams is to extend the textures out beyond the UV template boundary. Example of bleed on head texture shown below.

A "bleed" of 10 pixels for a 2K texture, and 20 pixels for a 4K texture is usually sufficient enough to get rid of those seams.  Texturing software such as Substance Painter or ZBrush, as well as other 3D software programs that have texturing capabilities usually have a setting to control the amount of bleed. 

I'm not sure that scaling the texture would be the optimal solution because then you'd run the risk of texture "landmarks" not falling in the right places. For example, nipples wouldn't land right on the torso, and eye, nose, and lip details may no longer land in the right place. 

I agree with everything said above. This is the same issue texture painting software runs into, but since we're tricking Poser into a texture painting role here, the usual affordances for working around the problem are not readily available.

Here's another test render with the HyperReal texture, this time after extending the texture, as DCArt puts it, in Gimp with the help of the erode function. So that's the postwork option I mentioned above, as opposed to the option of tweaking the conversion object produced by my script with the method so beautifully explained by FVerbaas.

Honestly, since I don't have that many textures lying around that I might want to convert to a different UV mapping, I'm quite happy with that solution for my personal use. Happy to do a bit of extra work though if anyone is keen on using this method in anger.

The script is on my github (same username as here) under the pydeltamesh repository just in case someone's curious. It's got the snappy name makeTextureTransferObject.py. Hit me up for the details on the process if you think it might be useful to you.

I am interested in the script and your process, odf, but I am about ten years behind many of you, in terms of technology and resources.  I guess that means I would like to understand your process, in hopes of adapting it for use with Poser Pro 2014.  I might be able to run a Python 3 script on one of my Linux installations, assuming the script is a standalone that doesn't require Poser 12.  Beyond that, I wonder whether this new UDIM business everyone seems to love nowadays is at all compatible with PPro14.

Will any of this work for a backwards user like myself, or am I too short to ride this roller coaster?

Coolio, I'll clean up the script a bit then and write down some instructions when that's done. No worries about the technologies. The script is standalone and should work fine with either Python 2 or 3, but I'll double-check that. I've used the same strategy for making Poser render out texture map back in 2010-ish, so your Poser Pro 2014 will do just fine.

UDIMs have been working nicely with Poser since at least Poser 6, probably earlier, although we didn't call them UDIMs then (at least I know I hadn't heard that term). I was told back then that some of the other tools that people liked to use didn't, which is why Antonia's original UVs are not using UDIMs. But no part of the process actually requires UV manipulation as such, so there wouldn't be a problem even if your Poser version couldn't handle UDIMs.

In the posts where various users have referred to UDIM, it seems like those long, composite textures  show up a lot.  If the process is combining all of the textures into one long strip, it seems like that would make texture editing and modification a hellish sort of process.  I guess I'm wondering whether those strip-images illustrate the staggering of overlapped UV sets or an actual merged texture-on-a-strip.

In the posts where various users have referred to UDIM, it seems like those long, composite textures  show up a lot.  If the process is combining all of the textures into one long strip, it seems like that would make texture editing and modification a hellish sort of process.  I guess I'm wondering whether those strip-images illustrate the staggering of overlapped UV sets or an actual merged texture-on-a-strip.

Good to know.  Thank you.

Script is looking good now, I think, and I've tested it with both Python 2 and 3. So I'll start writing up the process in a minute. In the meantime, it occurred to me that we can now also make V3 wear Antonia's textures.

So, that script... the full URL to the git repository is https://github.com/odf/pydeltamesh. Probably easiest to grab the whole thing as the script uses OBJ reading and writing functions that are in a different file. If you're not a git user, find the green "Code" dropdown and select "Download ZIP".

The script is called makeTextureTransferObject.py and it is in the subdirectory pydeltamesh - so that's two adjacent "pydeltamesh"s in the full path if you're counting. It is smart enough to find its dependencies, so you should be able to run it from anywhere without having to set the Python path.

It has four arguments. The first two are the target and source OBJs in that order, where target means the one with the UV sets you want to transfer to. Then come two specification files, also target first, then source, that tell the script which material zones belong to a common texture map. My specification file for AntoniaA looks something like this:

Body

  skinBody

  skinScalp

  nailsFingers

  nailsToes

Head

  skinFace

  lacrimals

  lips

  mouthInner

  tongue

  teeth

  brows

Lashes

  lashes

Now in Poser:

- Change the units to native Poser units.

- Create an empty scene with no lights. 

- Import the transfer object created by the script. In the import dialog, uncheck "centered".

- Switch to the front camera, then in the settings for the front camera, set all dolly and rotation values to 0, and set all scaling values to 100%.

- Set the render dimensions to your intended texture size, say 2048 by 2048.

With the preview set to texture shaded, I now see something like this:

Now it's time to start loading the source textures. In the material room, select the transfer object and for each of the material zones create a material that looks something like this:

Lights are off, so diffuse and specular shouldn't do anything, but I'm setting them to black anyway to be safe. :-) But the main thing is to plug the texture map into the alternate diffuse. Thanks to the material zone translation done by the script, the transfer object should have a single material zone for each texture map. Here I've loaded the arms and legs maps from the Toni texture, but not yet the body map.

Once all textures are loaded, go back to the pose room. It's now time to render out the individual texture maps. Start with the front camera DollyX set to 0, render, save the file, increase to 1, render and safe again, and so forth until all the maps are rendered. Here's what I see in the preview for DollyX = 1:

I can't say much about render settings. SuperFly seems to produce sharper output for me than FireFly, but that might just be my lack of FireFly skills. When in doubt, you could just try rendering at a higher resolution and scaling down the result later.

That's it for Poser. The final step is adding a texture bleed to get rid of the visible seams.

There are many methods for adding a texture bleed, so I'll just briefly describe how I do it in Blender. Similar methods may work in other 2d graphics programs, or one might use built-in functionality in a texture painting software.

0) While still in Poser, make sure to render on a transparent background and save as PNG so that there's an alpha channel.

1) Load the image into Blender.

2) Duplicate the layer.

3) Select the bottom layer and remove the alpha channel. Now the bottom layer has a white background, while the top one still has transparency.

4) Still in the bottom layer, apply the "Erode" filter a few times. This will "bleed" the texture into the white background. It will also destroy the texture, but we still have the original in the top layer.

5) Make sure the layer mode for top layer is set to "Normal", flatten the image and save it.

PS: If anyone feels the urge to take my script and turn it into a more user-friendly tool for Poser with fewer manual steps, please be my guest. Consider everything in pydeltamesh as being under the Apache License 2.0, even if it doesn't say so on the tin.

hborre posted at 8:45 AM Thu, 29 December 2022 - #4452676

@odf: did you mean "load into GIMP"?  You mentioned it earlier in the thread and I just wanted to verify that there is a mistake in your instructions.

The GNU image manipulation program is what I‘m using.