cakeller

I just noticed on the Tweet deck, Andrew says INTERSECTION IS NOW THERE!!! YAHOO!!

That actually makes sense, since 32 bit SHOULD (yes, SHOULD) NOT NEED A factor since it can hold exact 3D displacement info. so, question is, does using 8-bit mode with the factor work PROPERLY? Can it be used for now? and maybe confirm with Andrew if the 32-bit problem is a bug, or a workflow thing we're missing? (Sounds like a bug to me) [EDIT] RIGHT, I forgot about that (wasn't looking at it when I posted) ... Vector Displacement, Black NOT NORMALIZED is ABSOLUTE displacement distances (based on the space of course) I tested, it worked perfectly. So the question to Taros - does it work for you, and for that model? [/EDIT]

I think (please someone correct me if I'm wrong) but the depth export provides the current factor right after export. Did you use this factor, and it is coming out wrong?

PLEASE PLEASE PLEASE ADD THIS!! IMO, this is hugely important and would aid in detail work tremendously. Especially hard to reach areas. Please add an option to Zoom/Dolly/ whatever you want to call it... TOWARD the current pick point. To explain a little further, the current zoom (which is called "Scale or move forward/backward") only zooms toward the center of the screen. This is EXTREMELY TEDIOUS FOR DETAILS! This would improve navigation useability for me by 100,000,000 x (ok, maybe just 787,240.1 x... but... a lot!!!) I know bug fixes are top priority, but I feel like this is almost a pseudo-bug with the interface. This should be an option (IMHO) but is either missing, or I don't know where to find it. I thank you kindly, in advance, for your kind consideration of this pleading! PS> I just realized that the pan (move in screen space) also suffers from using the WRONG base point. PANNING SHOULD ALWAYS base it's screen based movement on the current pick point. If you move your mouse half way across the screen, the pick point should do the same. It is unacceptable for a 2 pixel move to send the model flying off screen. This can happen when the FOV is very small. BOTH OF THESE NEED TO BE ADDRESSED - PLEASE!

Good luck with your project. Sorry this didn't help. BTW, how much ram do you have? are you on 32 or 64 bit? if on 32 bit, do you have the windows /3g setting as recommended in the manual? Not suggesting you mess with this right now, but when your project is over, it could help narrow down the issue.

I was referring to programmatic decisions, not your choice/decision. I recently read an article on nVidia's web site that talked about using a 3D octree structure to store texture information (for a polygon mesh) instead of using a 2D map. What's interesting about it is that it only stores information where there's surface... so the data is potentially lighter than a 2D map of the same quality. So it is potentially possible that there could be a separate voxel grid or octree structure that holds the painted information, and this way, only the shader calls into that 3D octree color texture map. That way, there really is ZERO burden on the voxels to incorporate this feature. All the burden would be within the shader, and the density / complexity of the texture does NOT need to be 1:1 same density as the voxels. In fact it could be ANY power of 2 in size. (2x2x2, 4x4x4 ... 256x256x256) and since it's an octree structure, only leaves store data, so if there's nothing in a particular node, then it doesn't store any information. I could see it working... oh yeah, and this article was about it being GPU driven, so I'm guessing it'd be FAST-ish.

First off, I TOTALLY AGREE about the negative votes are just a bad idea. There are already two mechanisms to deal with obnoxious users - report them to admins, and/or block them. Neither of which I've ever had to do, but still it's there. With that said, I DO like the idea of being able to give someone kudos for an excellent explanation or good work or whatever. That to me is good thing, and helps herd qualify users, without the potential unhappiness that could result from the negatives. Just have a dig-it button without the hate-it button.

As I understand the UVs MUST exactly match between the high and low res mesh. If that's not easy / possible, you could import the high res mesh for per-pixel painting with NO SUBDIVISIONS and then go into the retopo room and import your 2k mesh (in the retopo menu, not file menu) this will bring your 20k mesh in with UVs, and allow you to wrap it around the high-res mesh. You can't do the big mesh option unless the UVs correspond 1:1 between your meshes. In fact, I don't think you can even do it if you smooth the UV baounds because there won't be a direct relationship between UVs and points. Let me know if that works or doesn't

Incidentally, while I was examining the custom navigation more carefully just now, I realized you can rotate around the screen axis instead of the vertical axis, and THAT'S AWESOME! (yeah I know it's been there, but now I get it) It makes it possible to get the limited "lighting" on areas that just isn't possible using the locked vertical axis type rotation. [EDIT:] by limited lighting, I mean because the shaders have some limitations, especially the pic materials. But the glossy ones really show the surfaces better for modelling, but only if you can get the reflection to slide across features in a few orientations. Which, now that I'm not locked to keeping the vertical axis vertical, YAY! it's awesome. [END EDIT] Now with that said, I couldn't find a way to ZOOM toward a picked point (you can spin around one but not zoom toward it) PLEASE could you add this or tell me where to find it if it's there already. I already made a post for this, but don't think Andrew got a chance to respond Thanks.

are you sure you can't just use the number pad camera shortcuts. Also, if you're looking for orthographic projection so that it's a direct side or front or top or whatever, Number Pad 5 toggles between ortho/perspective. As well, does the little cube icon in the upper right next to the camera drop down menu. as far as rotating the object (as opposed to the camera) using the transform tool, hold shift BEFORE selecting the rotation handle and it will snap to 45 degree increments. If this isn't what you're after, please clarify your question/request?

are you importing the big mesh as a reference mesh from the file import "big mesh" (which is for, importing those big ref meshes?

This is the thread that was already started for this topic. http://www.3d-coat.com/forum/index.php?showtopic=3767 Actually you could separate out portions of the voxels into separate layers using the hide - voxel->separate hidden. Keep in mind the hidden on the original voxels are still there. It's more of a copy hidden voxels to new layer. WORKS GREAT too.

And doesn't it take memory / processing to "decide" if I have to use the color version of voxels, or non-colored version? It might be a trivial amount of memory/processing, but... every decision a computer program has to make requires memory and time. Every OPTION a user CAN decide on has a cost (I don't know how big/small it would be) Direct X has a 3DTexture (3D Texels, which sounds to me JUST LIKE Voxels with color, cuz it is, or rather it's a 3D Volume Texture Map) class already there (but posts I've seen talk about it running out of memory at 512 x 256 x 128 for a size (which is SMALL) There's at least one voxel class library that was JUST released by Sony Pictures as open source, that includes the ability to have an arbitrary set of attributes for each voxel. it has limitations though (resolution at this point HAS to be a power of 2, eek and no random access to the entire structure, read it ALL or not. But in the works is being able to access sub-regions of the voxels. For 3DC and our uses, it seems like a pretty young tech (but DEVELOPING REALLY FAST thanks to Sony Pictures, nVidia, and 3DC and among others) Can't wait till it's fully possible and real time. But until then I'll do the layers thing. It IS a completely workable workflow for now. I'm sure it'll happen, just not this year... me thinks (me don't knows)

OK and here's what I think the Bas-relief tool could work like: the red is input value, the green is output value and the base is the distance along the locator from begin to end. and also an option to have the function through and through, or mirrored behind the begin point. Just my $0.02

Here's my comparison between pose vs barelief

ok, so I move the whole Bas Relief tool away from the object (like 100% away, not touching at all) Then apply... that's it? and so then basically all the noise in the void and noise in the solid will be eliminated / reduced? cool. I'll give it a try. EDIT: OK, I get it. The Bas-relief tool is just a scaling tool. In fact, I was able to achieve the exact same functionality with the pose tool in select with pen, using the box brush type to select. However, for obvious reasons it was more intuitive to use the pose tool. HOWEVER!!! I think you might consider an improvement to the Bas-relief tool as this: perhaps instead of a linear scale it is more of a gamma type curve, based on a mid-point. this would allow you to compress the background, and heighten the forground or vice versa. As a more robust alternative, perhaps it is a curves-like graph - in/out distances as percentage of the locators. that way you could do whatever cool arbitrary relief scheme you wanted. As it is, it ends up just a redundant tool. (as I mentioned I can do EXACTLY the same thing with the pose tool)

(and if you have more than one object you want to merge in from another source (maya max, or even objs from 3dc)... if you have more than one, tick the "merge to separate volumes" you'll get one voxel object for each sub-object in the obj file. Also if the mesh has _negative anywhere in the name it will cut out material from the current layer.

In the other viewers, they aren't smooth, they are rendered as IF they are smooth. If you look at the silhouette it's probably still faceted. Since the smoothing groups on OBJs / FBX doesn't actually affect the geometry, only the appearance of the shading on that geometry it won't affect the voxels. The Voxels are simply inside, or not inside the geometry. If you're hoping for a smoother representation of your mesh, you could use "subdivide" in the merge palette. But, smoothing groups - won't have affect in voxel-town. Someone correct me if I'm wrong (pretty sure I'm not though)

Actually, no the polygon cage is NOT a perfect representation of the ACTUAL voxel data. The voxel data is, essentially, 3D grayscale pixels. The polygon mesh is only the ISOSURFACE that defines the boundary along the 50% that boundary. have a look at this visual, it might help explain what I mean on the right is what I think MIGHT be happening when you toggle from voxel to surface and back, but not sure. DEFINITELY that IS hapening when you export to obj and back. On the left IS what's happening when working and smoothing, and sculpting. I'm not talking about the bubbles inside, I'm talking about the pockets of left behind noise from sculpting. which DOES go away when you export as OBJ, and re-import. Because the voxel pixels get reconstructed from the mesh. Also, the spell checker in the forum should really recognize the word "voxel" eheh.

When sculpting in voxels, since the surface is only the isosurface through the 50% value of solid/not solid. you could smooth a blob in space down so the surface dissappears, but the blob is still there, say at 45% or something. My question is, when you switch to surface, and then back to voxels, would that do the same thing as exporting the voxel as OBJ and reimporting? (creating the exact voxel field required for the surface, without any holes inside, or less dense pockets etc? The reason is, the less dense pockets of voxels (and super-dense pockets of void, or in other words the ALMOST surface) can affect the smoothing functions and make them unpredictable. hopefully this makes sense. I think this is why smoothing in surface mode seems to work more predictably. if toggling from voxel to surface an back to voxel cleans up the voxels, then YAY, if not I WANT this as a feature.

This isn't a great temporary fix, but it'll allow you to bend in an arbitrary albeit unintuitive way. the gear handle rotates around the axis... you can place that in any orientation. basically just make the gizmo's center axis your center of rotation. In some ways it's acutlaly very intuitive - to have an axis as a center of rotation

Yup, totally. You should be able to move the gizmo better, without affecting geometry. Right now you can only use the green handles do it However, it would be much more precise if the gizmo worked more like the gizmos for the primitive tools. Basically just add the XYZ move + rotate to the base handle. This could transform the coordinate space of the gizmo including rotation around it's own axis. To add to that, it would be nice to hae an XYZ drag handle to the end green sphere handle too, for more precision movement of all objects that have those green ball handles. (primitives etc)

Bump It! Heads Up Display style

Well, this doesn't mean it's not possible, but just consider, each voxel has a density. the surface is along the threshold of where the 50% solid/empty is. each 3D Pixel or Voxel, takes 1 Single Precision float (I expect that's what it takes since that's the highest degree in the rawvox format)... that's 4 BYTES / voxel. If you add a 32 bit color, now you're doubling the data to manipulate the voxel structure... and people already complain that they can't get enough resolution from the voxels and want MORE MORE MORE. IMO, the coloring doesn't need to be per/voxel, but maybe on a reference mesh? closest point on mesh from rendered voxel. But again, has to be fast - so maybe not? anyway, just my 0.02. I WOULD LOVE IT... but I'm thinkin' it's 3DC 4+ when hardware can REALLY REALLY handle it? I could be wrong about any of that... but not most of it Also, those shaders are HLSL (DirectX) and GLSL (OpenGL) ... perhaps a volume shader implementation. In fact, it'd be pretty neat if you could actually visualize the voxel field as the semi-solid it is, rather than just it's 50% isosurface.

Masking on voxels is actually KINDA possible with one tool. (unless you count the screen based masking, but that's not what you, or I, are talking about) POSE Tool! Supprisingly, or rather, NOT supprisingly, it's WAY MORE than just a pose tool. Change the method to selection by PEN, tick off Airbrush, or on for slower build-up of selection. you can smooth the selection too, and you can subtract from selection. The painted selection is fully 3D too, so it can be complicated to use/understand at times. HOWEVER it's a freakin' NEAT modelling tool! Check it out. My point isn't that the functionality that you want is there already, it's that the core functionality is already within 3DC... we just need it to be available for all tools!!! it's not just for being a poser.