V4.1 BETA (experimental 4.1.17D)

[Carlosan]

one thing is bug fixing and other is adding new features.

the first will give us a stable product and the second will give us a competitive product.

excuse me about my opinion... but...

a competitive product must be estable.

why ?

because when you offer new tool possibilities you are trying to gain a NEW users market.

so, how do you keep the new users attention ?

following the standard that another competitive product offer.

------------------------------------------

anyway, this is the 3D-Coat V4 Beta testing thread people... please use the post to focus finding new bugs

[BeatKitano]

excuse me about my opinion... but...

a competitive product must be estable.

why ?

because when you offer new tool possibilities you are trying to gain a NEW users market.

so, how do you keep the new users attention ?

following the standard that another competitive product offer.

I don't know about which competitive product you're talking about.

But if you're talking about zbrush I'll tell you this:

Zbrush is mostly know for its sculpting functionality, in fact before ztools, zbrush was totaly unknown with it's pixols, I'm probably one of the few hundreds that knew it before that and it was a toy no more less (and funnily this is what I often hear about 3dc because it looks "unpro" because "unreliable/not userfriendly" etc)

Now what 3dc ALREADY brings to the table zbrush doesn't in that area: unrestrained topology, without having to recalculate a projection to keep details, you edit on the fly "WYSIWYG". That feature alone could be the new gold of many studios IF it was stable, easy to pick for new users (no convoluted workflow or obscure names).

I can go into more details about what 3dcoat does better or even create new tools zbrush will probably have in 2-3 point releases, but the main difference between the two apps is that like mudbox 1.0: 3dcoat has the UI !

The UI the dreaded thing from zbrush from any user , you get to know it it's ok (you can even like it !) BUT it takes a while to get there. On the other end 3dcoat is standard, it's windows/palette based, not 3 meters long menus, dockable elements, this is true in every room.

Take the painting in zbrush: completely unintuitive at first ! 3dcoat: standard (I'm not saying easy because there's a need for more standard naming) and complete ! (spec in zbrush ? Nope. Bumps ? Nope, Easy acces to shading tools: try freezing your view with projection master first !).

Retopo: AHAHAHA I won't even try to go there, zbrush retopo tools suck and everyone knows it, most are using either topogun or 3dcoat for that alone ! Pixologic was so far behind they even llooked at 3dcoat to create their own autoretopo (which works well I gotta admit though)

Uving: well in this area zbrush has uvmaster, it's not bad, but it's limited.

etc

On that single thing alone if pilgways get it straight (goes with the v4 ui overhaul) it can get new users, no need for more functions (it can get more later ofc !).

What Zbrush makes me thinking of:

Amorphium (late 90's), first sculpting app available to general public. Terrible Ui. First of it's kind, subdivision based (but VERY limited) and software rendering no gpu acceleration. It Died.

Zbrush is like that software, very nice "first of it's kind" (to be usable in a pro environment not just hobby), but with a terrible Ui, subdivision based, vertex painting based, software rendering.

3dcoat is new, with a standard ui, it does sbudivision based somehow (voxel or sculpt room) AND freeform sculpting (voxel and liveclay) and it's gpu accelerated.

See the pattern, you just need to convince the people who actually use zbrush to try AND stay (the current issue, people aren't affraid to use it, they just get unexpected results).

[Andrew Shpagin]

Ok, as you suggested, started sculpting to feel it in action. Works good so far, found on little bug related to renaming of the instance. I hope it will be useful session.

This is what I started to sculpt (current WIP) - http://bit.ly/WzYHZJ

I will try to polish 3DC as well as possible.

[BeatKitano]

Not bad at all, in fact you could post here too

If you've got time, try to go into detail, most of the time it's when you try to get crisp detailing that you get unexpected behaviour (artefacts are coming after a few hours of work for instance very rarely at early stage).

Anyway I'm glad you try your software I'm sure you'll get more useful feedback that way.

[wave of light]

Just tried to merge a detailed surface mode head onto a body sculpt and I get this (going to open up in V3 and see if it's the same):

[Andrew Shpagin]

You should be careful with booleans in surface mode with heavy objects - it is very heavy operation that may fail. I hope it is undoable to make more attempt.

[gbball]

That is pretty cool Andrew! I wonder if you are creating these tools thinking of how they would be useful. Since you can obviously do it, maybe you can strike a balance between coding and sculpting, that way you'll spot workflow issues and potential bugs when certain tools are used in succession. Also, you'll be able to notice tools that aren't as useful.

Nice stuff!

-G

[Psmith]

It seems that there are several factors which decide the financial success and popularity of any computer graphic software offering. A software product can even succeed if it hits only 1 or 2 of these targets:

1) Ease of use.

2) Stability.

3) Quality of output.

4) Exciting features and functionality.

Taking earlier versions of Photoshop as a success model, we can see that it's success in the marketplace was due to it being easy to use, very stable, had excellent data output - and, lastly, continued to add new and exciting features with every release. Adobe hit every target of success with Photoshop.

If we take another product - say Maya - as an example of a successful product - we can conclude that it fails at being easy to use, only partially succeeds in the realm of stability, but provides excellent output data and continues to add new and exciting features over time.

When we look at the successful offering of ZBrush - we can readily agree that it is not easy to use, on the whole, but it is very very stable, it produces excellent data output and it continues to add new and exciting features to every major release.

If 3D-Coat is to succeed on a level that compares to any of the above software products, it needs to hit 2 to 3 of these success targets to become and remain competitive in the computer graphic software marketplace.

P.S. However, none of the examples could really have become successful if they had not hit target No. 3 - Quality of output. Gotta have it.

Greg Smith

[BeatKitano]

It got 4, I would say 2 is on the way, 1 is definitely the worst offender after 2 and probably the easiest to reach with few changes.

Btw fixed:

If we take another product - say Maya - as an obsolete example of a successful product - we can conclude that it fails at being easy to use (I disagree but I'm biased with my history with alias), only partially succeeds in the realm of stability, but provides excellent output data and continues stopped to add new and exciting features over time.

[Psmith]

What about 3?

Greg Smith

[BeatKitano]

That's another point I would gladly discuss when the other points are at an acceptable level, for now I don't want to hurt morale (even if there's good things).

Like zbrush 3dcoat is a content creation tool and can be used in way which doesn't always involve maps (that's what is not always optimum geo's mostly ok).

[Psmith]

Maya is a good example of something else regarding product success and failure:

1) It had a fantastic run up to its peak of success.

2) Could not sustain its market share

3) Got absorbed by an industry Giant

4) Got completely digested by that Giant

We will continue to see the contraction, absorption, and complete digestion of many more computer graphic software products and the companies that spawned them. (We'll see what happens with Modo, next).

Greg Smith

[Tony Nemo]

With Andrew's comment, "You should be careful with booleans in surface mode with heavy objects - it is very heavy operation that may fail. I hope it is undoable to make more attempt." I recall my attempt to subtract one layer (in voxels) from another layer ( 7x in surface mode) and after an interminable period the program crashed. When I attempted the same operation with both layers in voxel mode, it was prompt and successful.

It will no doubt take time but I'm sure 3DC will become idiot proof.

[BeatKitano]

I wouldn't say it's idiotic, but obviously using two voxels layer is easier to manage computation-wise (if they're both in memory before, there's "less" chance it crashes after the operation), especialy if you try to use a surface with that high level of subdiv, back to voxel it needs a really powerful machine to handle it.

What would be best maybe is to warn the user for heavy/risky computation like this. Zbrush does it (maybe a bit too much but at least you're not surprised if there's an issue) now from an user perspective, this shouldn't be an automated warning (meaning you get it in the face for the lighter operation) but when the computation required is high a warning would be cool. (btw this is exactly the kind of situation I was freaking about with decompose + floating bits)

Maybe a simple

Risk x blablalba

Do you want to proceed ?

[ ]Don't ask me again for this session

-Yes -No

[wave of light]

You should be careful with booleans in surface mode with heavy objects - it is very heavy operation that may fail. I hope it is undoable to make more attempt.

Yep, it's present in V3 too. I have plenty of backups with incremental file names, so I could go back to a lesser detailed model and try again. I was instructed to do it all as one voxel and then go to surface mode for detail, but I wanted to test the brick like details on the head first. Oh well, it's all part of the learning curve.

With Andrew's comment, "You should be careful with booleans in surface mode with heavy objects - it is very heavy operation that may fail. I hope it is undoable to make more attempt." I recall my attempt to subtract one layer (in voxels) from another layer ( 7x in surface mode) and after an interminable period the program crashed. When I attempted the same operation with both layers in voxel mode, it was prompt and successful.

It will no doubt take time but I'm sure 3DC will become idiot proof.

I use 3DC mainly for texture maps, UVs, etc. And for ages wanted to go back and finally nail the sculpting side of it. I will go and see if I can return the head back to voxels without too much loss of detail, and then try to re-merge.

I know what you mean by idiot proof. I'm very good with software, I can pick stuff up easily. But this has been a steep learning curve already, and mainly because 1) I'm using a the BETA version, 2) stuff that looks like it's messed up when really it hasn't e.g. 'show voxels in paint room' and 3) try out my own workflow, with guidance from other members, only to find issues later on.

For me, 3DCoat is a very detailed piece of software, with of lot of options, and different ways of reaching the same goal. If you miss one, or don't know about it, it can sometimes look like the software isn't working, whereas it's usually the user.

[stusutcliffe]

I had an awful time trying to merge/move my collar to torso on my "Riddler" model, either one or the other would just vanish! So I made them both Voxels, It worked straight away! Though, to keep the detail I made the voxel transfer really huge then reduced it when back in Surface mode.

[BeatKitano]

I try to avoid surface<>voxel conversions. It's not as bad as it used to be but that's another way to get destroyed meshes. Also doing those conversion often are a good way to trigger the famous artefacts around the mesh.

[wave of light]

Ok, wow. I turned the head back into a voxel sculpt and up'd the required polys to 6 million to try and keep some of the detail. Then I merged it with the voxel body... and waited and waited. 10 mins later and I get the merged head and body, except its 60 million polys and my PC is having a stroke! Every time I move the mouse I have to wait 3 mins for it to settle. Got to the decimate command and now it's stuck at minus (yes minus) 14%.

Whoops.

[BeatKitano]

Yep, x7 is pretty high for voxels, and it's completely dependant on the size of your model. You can try resizing your model a bit before trying again, that can help.

One thing that I often miss with heavy process in software is a cancel action, I know it's tricky to add sometimes because you need to make multiple threads or asynchronious processing but the ability to cancel an operation that look like is failing is priceless imho.

Something to think about for later maybe ?

[wave of light]

Yep, x7 is pretty high for voxels, and it's completely dependant on the size of your model. You can try resizing your model a bit before trying again, that can help.

One thing that I often miss with heavy process in software is a cancel action, I know it's tricky to add sometimes because you need to make multiple threads or asynchronious processing but the ability to cancel an operation that look like is failing is priceless imho.

Something to think about for later maybe ?

Sorry, this seems to have turned into the 'help Ricky sculpt better' thread. Giving it another go as we speak.

Yep, a cancel button wouldn't be a bad idea. However, all these hints and tips are helping, thank you.... but if I were doing this trying to hit a deadline, I would have jumped out the window by now!

[Tony Nemo]

I noticed that the size went from 7x to 5x on returning to voxels from surface mode. (for safety sake this is being done in 3.7.18F)

[AbnRanger]

Latest BETA4 was really essentially unstable in retopo room and I found it only after getting back from vacation... that is really not good, especially for Linux user who have no BETA3...

Sometimes I making some real models in 3DC and it always helps a lot. Maybe I will try to do something again, Raul will try to model something real too. In recent build all what is done is bugfixes + one little improvement of materials + Raul's tools. I plan to continue stabilizing. Generally Mantis helps a lot.

Don, you often tell about CUDA but I don't do this request for years because it will require much more efforts and time than will bring back efficiency. All I want now from CUDA is don't crash because it may just crash unpredictably without any visible reason. Maybe someday I will get idea how to do CUDA support better.

When users have CUDA 4+ installed and 3D Coat has only been compiled for CUDA 1, it's no wonder there are conflicts. Most developers that utilize CUDA, will keep it up to date, in order to take advantage of the numerous improvements with each major release. CUDA is stable in other applications (Octane, iRay in 3ds Max, Cycles for Blender, etc.), why not 3D Coat?. There is a Visual Profiler in the CUDA Toolkit that analyzes problems and bottlenecks, and offers best case solutions.

Another option might be to switch to OpenACC, which is a high level (no need to rebuild your code from the ground up) standard that exists to greatly simplify CPU multi-threading to GPU multi-threading, with directives or hints in your current code. Practically no compiling involved. A lot of major corporations collaborated on that new platform....including Cray, PGI, and NVidia. Could be worth a little research....no? Plus, there would not be any need to have CUDA and non-CUDA versions of 3D Coat, and users wouldn't have to rely on NVidia cards only.

[AbnRanger]

Ok, as you suggested, started sculpting to feel it in action. Works good so far, found on little bug related to renaming of the instance. I hope it will be useful session.

This is what I started to sculpt (current WIP) - http://bit.ly/WzYHZJ

I will try to polish 3DC as well as possible.

Nice model so far, Andrew. Try using the Pose tool on a 3-10mill+ poly object. Paint/Brush select an area (see how slow it is) > use the FFD > select some control points > deform with gizmo. Pretty slow. I know you can cache the layers to use on proxies, but there are many occasions where this is not a feasible workaround. Such as deforming a lot of layers/object via the "Through All Layers" option. Sometimes it works and sometimes you get a mess.

Try using the Pose Tool on the Mannequin, and make sure to use "CHOOSE BY OBJECT." Then deform it. Try clicking CLEAR SELECTION. I believe I already reported this stuff a few months ago.

Try merging a mesh, modeled externally, and getting rid of the faceting without degrading all the detail it has on the model. Try baking/merging that football glove I sent you (the one that has different sub layers and with some Voxel Painting. Should be loads of fun.

[AbnRanger]

That is pretty cool Andrew! I wonder if you are creating these tools thinking of how they would be useful. Since you can obviously do it, maybe you can strike a balance between coding and sculpting, that way you'll spot workflow issues and potential bugs when certain tools are used in succession. Also, you'll be able to notice tools that aren't as useful.

Nice stuff!

-G

+1. Retopo the models and if they are nice, you can use them as Primitives/Preset models in the Models Pallet. Also, the Mannequin and the other Human figure should be scaled the same size. One opens up HUGE and the other, not so much. Should be a little bit consistent, otherwise it reflects poorly on the application.

[Andrew Shpagin]

Ok, there i my WIP and report - http://3d-coat.com/forum/index.php?showtopic=10640

[wave of light]

Ok, kinda nuts, but this seems to have worked. I jumped out of S mode with the high detailed face and set the poly count to 6 mill. Then I reduce that and the body in size using the Transform tool. [x88] in the VoxTree menu. then merged. And waited. Resulting 3 Mill voxel model. Yes I have lost detail in the face, but at least I can continue on with the rest of the sculpt and then finally, when I am happy with the pose and low-mid detail I can start on surface mode and re-introduce the cracks, etc. Have noticed that all the brushes are a lot slower now, unless I use smaller sized brushes.

My question remains, how would this work in a production pipeline... it would be a nightmare if the TD wanted big changes to the pose or additional features added to the model.

[Andrew Shpagin]

BETA5 for Linux uploaded

[philnolan3d]

Ok, there i my WIP and report - http://3d-coat.com/f...showtopic=10640

Not tired of dealing with bugs yet I see.

[Tser]

Andrew, good to see that you are tasting your own medicine, this is the only way to get direct feedback from your own creation, I really hope you utilize your whole application and not just sculpting and vertex painting because you have to go through the entire work flow from sculpting to exporting your final retopo-ed, low poly mesh with its Textures and Displacement maps for use in another application.

[Tser]

I have noticed a strange happening when starting/running 3DCoat in Linux, all the other OGL programs which are currently running become unstable and eventually crash, this never happens if 3DCoat is not running. Also, my video card fan starts to up its rpm only while 3DCoat is running, even though I am NOT running the CUDA version, if I run the CUDA version my video card sounds like a leaf blower (Quadro 4000 running the 290.03 driver, OGL version 4.2 ).

T.