Dead or Alive series formats and tools

[Lilstormcloud]

Yup, sure can.

[Rosalin]

VR suit is a gold mine.
It has really good potential.
I'm workin' on it.

Well.. this has some troubles as well as it's potentials.
***

lil,
That looks good.
Sounds complicated

[Lilstormcloud]

where can i find some bunnies

[Protocol X27]

LSC, the idea you had earlier about using the UV coordinates to match stuff was genius!!

I started testing two theories this weekend. The first is matching two different body vert sets based on UV, a Lisa to Christie check found 5,000+ matches. I rewrote the verts from one to the other. The only problem that the UV is mirrored so I need to distinguish the correct x-axis value when doing the re-write. It worked perfectly on a non-mirrored element like toe nails. Since I don't understand hex numbers and negatives well enough, is there a way to see which vert is above or below the X-axis?

The second thing I tested was related to that alpha problem. I Zero'd out the last three values (24 bytes) for each vertex in an object in question and it didn't appear to have any affect on the alpha value. I have a thought that it might be possible to drop a 44 object down to a 40 and see if that does anything. The objects with dual layers seem to already be untiled and ignore the alpha layer OR use it for another purpose. Perhaps if I kick it out of the group that is using the alt texture, maybe I can pull something off.

[Lilstormcloud]

if you search IEEE754 there'll be documentation

[Protocol X27]

You rock! That's what I needed.

They are indeed floats. I mapped a Lisa costume to a Christie Body and it matched about 75% of the verts. The only stuff that didn't was the wrists, feet and miscellaneous bikini depression areas. Even if not perfect, it's a ridiculous time saver. I'll do a test on Marie Rose soon to see how it really works.

[Rosalin]

Wow... that worked.

I think there's gonna be no problem if you use just transmographer to do it after using script. without identifying floats are below zero or not. only if that is mirrored mesh.

Thinking about toes and nails again... it had been possible to distinguish them by vertex x or y coordinate lol.

********************

111.png

Applying MLAA to screen shot.

1. Download Pixel bender as Photoshop plug-in or Standalone, whatever you want.
http://www.adobe.com/devnet/pixelbender.html

2. Download MLAA for pixel bender.
https://www.adobe.com/cfusion/exchange/ ... id=2230024

3. Run Pixel bender and open the screenshot you want as PNG format and File>Open filter choose MLAA file, and Build>run

4. 0.1, 0, 255 (default?) will be fine.
save it.

5. In stand-alone version, the canvas size might have been expended unwanted.
just load original image in Photoshop and copy over the MLAA filtered image as a layer(not as smart object). that will simply handle the canvas size.

6. After you applied the MLAA filter, the highlight in the eyeball has been a little bit blurred. The highlight in the eye ball is better when it's sharpened. You can simply recover it by erasing that part when the MLAA filtered image's layer is over the original image layer.

[Protocol X27]

Ros, great tips on the image resolution. Especially for the multi-tiled concept. It never occurred to me to do something simple and piece them all together.

Vertex Snap Update: So essentially Lisa > Christie worked mostly, Lisa > Marie Rose will require more adjusting. Lots of parts like neck, butt, torso and others will have issues, but its still less then the usual amount.

Tomorrow I'll work on bundling the snapper and the fix vertex buffer stuff which I forgot to do.

LSC, Hopefully, I'm not working something that will invalidate what you're working on. Is there anything you would want me to focus on for vertex writing. I think I should be able to grab any part of the buffer and swap it, based on current method.

[Lilstormcloud]

nice find on the photoshop plugin.

[Rosalin]

lil,
For me it looks too advanced to make that process into scripts.
I didn't yet understood any of concepts of inserting part of mesh to another character's TMC & TMCL.
It's just looking that some process of adjusting 'after' inserting meshes & verts.

**
Anyway, if vertex snapper script works, it's gonna be really easier to swap costumes to another's one.
(ex. Azurie to MR, lol)
I think partial morphing should be quite useful for repairing some unwanted modified parts after using the script.
I'll explain it later.

**
Yeh, mlaa filter is quite nice. but it could be looking a bit blurry then the original image.
original(pixel jaggy) vs mlaa filtered(blurry)
it's your choice.

And it can't do quit well with 1pixel thin hair lines. (ex mr)
not like SMAA.

[Lilstormcloud]

It's all good.

[Protocol X27]

Ok, apps packaged.

I may be out of commission for new tech for a couple days on account of some real world stuff.

Ros, I messed around with Gimp a bit and its Filters to remove some of the anti-aliasing. One called Van Gogh seemed to help a bit. I'd run that a couple of times and then sharpen. It's a big wonky, but just another method idea.

Do you feel ready to tackle injecting? Do you feel comfortable with finding vertex and index groups for specific mesh pieces? If so we can start explaining some of that methodology.

LSC, what are you doing to approximate x,y,z? Are you doing that based on an injected mesh and comparing against the source one?

I think my next small app project will be to do an auto neck thing for the two top rings. I think that if all verts were sorted by Z(?) value, I could grab the top 20(ish) verts, compare to their source and then copy over the buffer values.

After that, I really need to get the auto injector moving. I can do an injection in 15 minutes, but it's becoming a lot repetition that needs to get cut out.

[Lilstormcloud]

the rounding take care of minor variances

[Rosalin]

I tried Body snapper, but unfortunately, it seems to damage source file.

I tried it for lisa004 to same copy of lisa004 but source also damaged and couldn't be opened in 3dsmax and noesis.

Is there any fine samples for testing? I tried to pick some more, but I couldn't open the results.

And thanks anyway.

**

For about mesh injecting, I hope that I can keep it for later.

I've been quite rushing and I feel a little exhausted about understanding concepts, especially with HEXes.
I want to concentrate on doing something with what I know, than learning what I should know, currently.

[Protocol X27]

LSC, thanks, the rounding was what I'd been wondering about since 3dsmax only has a 3 decimal place threshold. Are you grabbing any data beside the bone weight & id. Just curious since I think I've done that w/ body shop for other mesh pieces although not the mass copy stuff.

Ros, i mentioned in the instructions that it doesn't use an actual tmc, just a hex file of the verts witha .tmc extension. I did that since it will eventually implement, tmc reading, and prevent slowing it down when processing a text file. You can find the body verts by using the obj properties info found in 3dsmax. start & length values need to be converted to hex, save the vertex values to a generic file. Once updated, paste vert data back into the tmcl that they came from.