PDA

View Full Version : Talked to Steve


My Fault
02-07-2003, 02:27 AM
OK, so I spoke to Steve today and here's the general gist of my conversation with him.

First off I have to say HUGE thanks to Steve for taking the time out to give me a yell. Even after some of the asinine email he's received from the more emotionally challenged users in the last few weeks he's still pretty upbeat and positive. He also sends out a big thanks to all the people who have been sending him ideas, bugs (with data), etc. The support is very much appreciated by everyone at Hash central.

Renderer
They are definitely taking a big stab at the renderer, working on many of the issues we've all talked about. There will be more options available for controlling antialiasing and creasing issues and increasing the renderer's speed. It's still very early but it's good to know they are listening and trying very hard to get it worked out.

Stability
Bug fixing and stability is at the top of their list, occupying around 70% of the programmers time right now. As for bugs he asks that if you do have any please email him directly and include any and all data. The chances of bugs getting squashed goes up exponentially when they have the project file to work with.

Subdivision Surfaces
SubD's are not going to happen. Steve mentioned "Feature Freeze" a few times in our conversation and the amount of work to implement SubD's is too great. They feel their patch technology is very powerful (see the other thread on Martin's email as to some of their reasons why) and are working to solve issues we are having such as creases. I must admit I was disappointed to hear this at first but I do also agree that the pluses of patches outweigh the negatives. I still think there's room for additional modelling tools either from Hash or by the user community through the SDK to make modeling even better in AM.

Well gotta head out for a bit. I'll add more later.
:wavey:

Wegg
02-07-2003, 02:40 AM
I don't see anything wrong with the patches themselves. Infact I love modeling with patches. The issue is more with the creases. I think we need to start looking at ways of showing the Hash guys what we mean instead of leaving it up to "subjective" opinions.

Show how easy it is to create a smooth organic shape in a SubD environment and show how it is that much harder to achieve the same level of smoothness with a patch.

What do you guys think?

Hookflash
02-07-2003, 02:45 AM
My Fault: Ok, first of all, are you calling us "emotionally challenged" because we complain after feeling like we've been ripped off? Secondly, why should I (or anyone else) have to go through all the *work* of sending bug-reports manually? Frankly, it just takes up too much of my time. I want an automated system (or even a web form), that will gather relevant information and allow me to give a short description of the problem, attach files, and click "Submit". Also, it would be nice (for the developers) if A:M would write out some sort of error log when it crashed. Finally, I think you are wrong about the pluses of patches outweighing the negatives, and it seems that most of the more experienced 3d artists (with the exception of a few Hash fanatics) agree with me. SubD's are becoming the standard for a reason. Just my $0.02.

My Fault
02-07-2003, 03:00 AM
Originally posted by Hookflash
Just my $0.02.

I want my change back :p

Nonproductive
02-07-2003, 03:02 AM
Hookflash - I seriously doubt anyone thinks patches are better then Sub-D's in general. However, to many folks - patches are not a "show stopper." Beyond the creasing issues they are nearly as capable as sub-d's and some people actually *like* spline modeling :eek:

I think pushing for sub-d's is wasted energy at this point. Not because it's a bad idea, simply because it's a lot of resources to commit to adding a feature that's not "needed." (needed being subjective I know...)

I agree with you on the Bug Database. Not only is it easier for the user - it also means Steve can spend less time weeding through emails and answering phone calls. Additionally, it gives the users some place to go to and find out if what they are experienceing is a bug or not - or at least if anyone else is having the same issue. I think it makes sense on many levels to implement something like this.

To me it looks like Steve discussed the "Big 3" that I recall reading here (and commenting on): Renderer, Stability, Creasing. In my mind, if they get those 3 "beat" then they have made huge strides (until the enxt code rewrite). My question though is whether or not this will be v10.5, v11 or a more typical "we're working on it" and stretch it over several revisions at $99 a pop.

I still think that they need to add *support* for OBJ import/export with the whatever caveats they want to attach to it. But have *some* method of getting models in and out of the program.

In any event, Martin and Steve have done a lot of talking the last few days - now we can wait and see how it pans out. I'm anxious, to be honest, but I will *not* spend another $100 until at least the "Big 3" are *done*.

My Fault
02-07-2003, 03:15 AM
Originally posted by Wegg
I don't see anything wrong with the patches themselves. Infact I love modeling with patches. The issue is more with the creases. I think we need to start looking at ways of showing the Hash guys what we mean instead of leaving it up to "subjective" opinions.

Show how easy it is to create a smooth organic shape in a SubD environment and show how it is that much harder to achieve the same level of smoothness with a patch.

What do you guys think?

I think it's a solid idea. I know there has been a lot of behind the scenes talk with Hash about this issue but I think good solid examples are always a good thing.

My Fault
02-07-2003, 03:32 AM
Originally posted by Nonproductive
I think pushing for sub-d's is wasted energy at this point. Not because it's a bad idea, simply because it's a lot of resources to commit to adding a feature that's not "needed." (needed being subjective I know...)

That's really the big point Steve was making. Personally I think Sub-D's will show up eventually but at this time it is really low on the feature list and the amount of work it would take to implement is not there right now. If they give us the ability to control the creasing issue that will solve the majority of problems users are having.

johnathan
02-07-2003, 03:40 AM
as i'm not taking this all as gospel, count me among the ranks of the emotionally unstable. that being said, if the 'big three' get whacked, then i'll be a happy hasher.

the big question in my mind, as nonproductive says, is how many nonfunctional years will we have to endure... and pay for?

-jon

bugzilla
02-07-2003, 03:52 AM
I think this is GREAT news. I am so glad to hear that Hash is focusing on getting their existing tools working and stable than adding new features. What Hash needs first is a ral renderer that doesn't take a lifetime to crank out a reflective sphere sitting on a bump mapped plane.

At present I own C4D R8 and Lightwave 7.5 plus downloaded Wings 3D and have used Maya and Max at work. SubDs are indeed awesome but since I started with splines, I really have a hard time working in anything else. It's like getting into a bad habit. Still, if Hash can fix the creasing problem splines are still the fastest way to model organic creatures.

JTalbotski
02-07-2003, 03:57 AM
My Fault,

Can I take it that better hair with guide hairs was not discussed? Or does the feature freeze cover that?:hmm:

My Fault
02-07-2003, 04:02 AM
Originally posted by JTalbotski
My Fault,

Can I take it that better hair with guide hairs was not discussed? Or does the feature freeze cover that?:hmm:

No we didn't get to it, sorry. You should just give Steve a call Jim I'm sure he'd love to hear from you.

HellBorn
02-07-2003, 08:29 AM
Sounds god except for the none SubDs.
I do understand that it's lot of work but..

It would of course be a greate breaktrough to the spline modeling in AM if the creases was removed, but how much less work would that really be if it's to be solved in a well working way and not something like the crap porcelain material.

Also, the SubDs is not only about the creasing. It's also about opening up AM for decent model import. If there will be no support for SubD then the time it takes to write a miracle importer that can convert a polygon model to a spline model has to be added. And it has to be so god that there wont have to be any or much postwork inside AM. It probably would have to get imformation from the user in the process so that It knows where joints will be added and the direction of animation so it can place the splines in an optimal way. It would have to increase density where needed etc etc. Untill that is done there is no model import worth the name. If Hash want AM to stay splines only I really think this is neccesary in a world going SubD.

SubDs are indeed awesome but since I started with splines, I really have a hard time working in anything else. It's like getting into a bad habit. Still, if Hash can fix the creasing problem splines are still the fastest way to model organic creatures.

Well it's probably a bad habbit as you say yourself. ;)

I do organics a lot faster using SubD. With splines, even with the creasing fixed, you still would have to do a lot of planning before starting to model, or it won't be a very god model to animate.

I do understand Hash cant fix everything at once so:

1. Freeze the features and bring us stability.

2. Get the new render going.

3. Give us SubD or at least a proper support for polygon models or give us that miracle importer.

4. Add whatever features they want.

Nonproductive
02-07-2003, 01:50 PM
Hellborn, your points are valid to an extent but again, I really think the Sub-D topic is dead.

You can read my other posts here - I am not, at present, a fan of Hash. However, the plain truth is that if you don't like splines - don't use A:M. It's really that simple. Splines are the foundation on which all of A:M is built and they are not going to go away...ever. Sub-D's are also not something A:M *has to have.* They would be very nice - since the animation tools are so sweet and many people would love to take advantage of those tools in conjuntion with their sub-d modeling skills. They are not needed though. You either accept splines in A:M or you use a dedicated app like Messiah or Kaydara Motion Builder with a package like Lightwave, Maya or C4D with poly models.

I also disagree to a certain extent with your description of what an importer has to be. Because slpines are so sensitive to placement, bias, and direction I don't *want* an importer that makes those decisions for me. Give me the ability to get a low poly cage into A:M and I can rebuild it with splines by building my splines on the existing points - or, if it's just going to be a prop I can tweak the cage to get it into a decent shape. Because splines need so many fewer points to define the surface this can be done pretty quickly. When you start letting the importer make modeling decisions it gets funky.

I doubt you will see anything beyond basic poly model import for a long time to come. If that's a "deal breaker" for you then you can probably safely move on now. If you accept the fact that to use A:M you have to embrace spline modeling then there seems like a chance that the rest of the community's valid complaints will be addressed in the future - hopefully before we all grow too old.

JoeW
02-07-2003, 05:21 PM
While I agree that subD's are sweet, and I like modeling with them, I still think that splines have a place in my modeling toolbox. Splines let me "sketch" in 3d space. I don't think they require any more planning than subD's.

Splines will always yeild a much more efficient mesh - you can get the same shape with a less geometry, and that's very nice when it comes to smartskinning and rigging.

What *I* want is a spline patch surface that RENDERS as if it were sub'Ds.

I think the coolest tool that could be added to AM's modeling arsenal is the patch equivalent of LW's "spin quads". For those of you who don't use LW, what it allows you to do is pick neighboring quads (patches) and by hitting Ctrl-K you "spin" them, or change the direction that the edges are going. Once you get used to spin-quads and bandsaw, it's very hard to live without them.... hmmm... maybe I'll have to talk to Tony Jones and see if he thinks that's a possible tool......

Here's a little image that shows what it does...

ftp://ftp.hash.com/users/joewllms/samples/spinquads.jpg

JoeW

Goon
02-07-2003, 06:16 PM
Even if people like splines so much, couldn't they change the way the surface is treated? My understanding is that it is currently a quilt of curved surfaces which is basically stitched together at the edges. What if it became a continuous surface, similar to Sub-D, with the vertexes not quite attached to the surface, but very, very close. This would retain the closer, tactile control of patches, but would have a seamless, uncreased surface.

JoeCosman
02-07-2003, 07:02 PM
Originally posted by Goon
Even if people like splines so much, couldn't they change the way the surface is treated? My understanding is that it is currently a quilt of curved surfaces which is basically stitched together at the edges. What if it became a continuous surface, similar to Sub-D, with the vertexes not quite attached to the surface, but very, very close. This would retain the closer, tactile control of patches, but would have a seamless, uncreased surface.

as a matter of fact, yes. and this is where Martin needs to pay attention(yes, I'm telling you how to code)

The renderer uses polygons. we can all admit that now, no need to keep hiding it. how the patches get turned into polygons is what causes the creasing.

take, for example, a face made in AM. render it out and look at the ghastly creasing

now export that face with 4 or 16 subdivisions and import it into Lightwave to render.

why does it look smoother??

it has to do with how normals are calculated and hash's tesselation method.

you see, if you tesselate the surface and THEN calculate the normals on the polygons that cross the creasy areas, then you get creases in the rendering.

if you calculate the normals, THEN tesselate, you get smooth no matter what.

Each patch is separate with matched tangencies, so grid-like meshes look smoother than, say, a mesh with a giant 5-pointer in it. Hooks match tangency to the neighboring points. Five pointers are basically 2 quad patches and a tri that are stuffed into the hole. hence you get a large crease from the center out in non-uniform, non-flat 5-pointers.

By forcing each patch to have a single normal vector before rendering, you can effectively ignore anything inside the 5 pointers, and tri-patches(which will appear to render smooth.)

This is exactly how the porcelain material works. Except the current method also ignores gamma and magnitude, destroying any curved shading inside the patch, making your model look like it was made with a game engine instead.

tesselating the mesh to match the patch curves, then shading it using the tesselated normals does not seem like a good solution. 15 years of renders with Hash products prove this.

the ideal solution would be like the porcelain, but using a subdivision level to calculate normals on. so if it's set to 4, then each patch gets sixteen normal vector checks(one for each tesselated face). if it's 1, then it gets 1 normal check(one for the whole face), which looks just like porcelain. if it's 16, then 256 normal checks(one for each tesselated face)

the advantage to this is that you can have 5s and 3s in the same mesh without getting any creasing. The normals force the tessselated mesh smooth.

lemme know if I got anything wrong here.

-joe

JTalbotski
02-07-2003, 07:46 PM
Originally posted by JoeCosman

lemme know if I got anything wrong here.

-joe

Joe,

You're slipping, you forgot to call somebody a nasty name. Some people might think you like Hash.:p

(Actually, that was one of the best posts, right or wrong, on this forum.)

My Fault
02-07-2003, 07:57 PM
I know they are working on something that sounds very similar to what you posted Joe (not sure if it is exactly that though, me am no 3d programmer, it hurt me brain ;) ). Not sure where it will go but it sounds promising.

And Jim's right, unless you say the word "beeyatch", someone may take you seriously :bounce:

JoeW
02-07-2003, 08:03 PM
I agree with what you're saying - except for one little niggle - a 5-point patch is actually a set of 5, 4-point patches that terminate in a center point, and use kind of "hooks" into the surrounding splines - kind of like the petals of a flower.

If you make a flat 5-point patch and lift one of the vertices out of the plane of the others, the underlying structure becomes pretty obvious. If there was a way to control that "virtual" center point, 5-point patches would be easier to control....

But I'm sticking to my "model with patches, render like SDS" story..... ;)

JoeW

JoeCosman
02-07-2003, 08:26 PM
I guess it depends on what version you use at the moment.

several flavors of 8.5 and 7 would export 5s as two 4s and a 3, and others do the 4's terminated at the middle.

I do recall in several builds, they switched between using a pinched four, and a 3-quad for the tri-patches.

either way, they get lousy results. and exporting from 8.5 uses 5's, 3's and hook patches as separate meshes.

pretty nasty to clean up.


oh yeah, shiznit, beeeyotch, nipple, and *scoff* just for good measure :-) :-)

-joe

JoeW
02-07-2003, 10:03 PM
Originally posted by JoeCosman

either way, they get lousy results. and exporting from 8.5 uses 5's, 3's and hook patches as separate meshes.

pretty nasty to clean up.


True.... true.....

Hopefully, the way that surfaces are sampled will change, and this will all be a distant memory.... sigh.

JoeW

Wegg
02-07-2003, 10:35 PM
I also got a little miffed at the way 5's handled decals and procedurals. It seems the UVs for 5 pointers just don't work quite right and by doing so, draw they eye to those areas.

No I can't really show an example because I'm not using AM for much of anything lately but. . . it was something that really bugged me at the time.

John Keates
02-07-2003, 11:27 PM
If they can sort out creases in the way suggested then I will be a happy boy indeed. :drool: I am having a hard time getting my head round sub ds and I realy like drawing my geometry. Lets hope that it gets in for v10.5. If it does then I will possibly forgive Hash for wasting my time with v9. :thumbsup:

ypoissant
02-08-2003, 06:16 AM
Joe C,

This is an interesting analysis you do there. I've played with the idea and came to some contradictions. I'd like to add some observations.

Subdividing a patch into 16 patches only subdivides each spline segment into 4. 4 straight segments is not enough to snugly match the true curve and this is why you get smoother renders when you export to polys.

If you tesselate fine enough to snugly fit the true curve, then you will get the crease again. If you tesselate coarse enough that you jump over the crease area then you will never see the crease. Not because it's not there but because you jumped over it. And this is exactly what happen when you convert to poly from Hash. 16 division is coarse enough that it jumps over the crease. Just repeated your analysis using other words.

I would say that subdividing into 4 (16 normals) would be the finest you could attempt without having the creases obvious again. But the creases would still be there though. They would only be so blurred out that they would be unnoticeable. At 16 subdivs (256 normals) the creases would be easily noticeable. (This said, I can guarantee you that I could devise a set of patches that would show the creases even at 4 subdivisions).

So now we have a sampling mechanism which is designed explicitly to jump over gamma tweaks. Because this is what the creases are all about. They are default gamma settings that we don't like. But what if, as a modeler, I would like to tweak the gamma in order to obtain a specific curvature but the sampling mechanism is blind to it?

With such a coarse normal sampling, designed specifically to smooth out gamma, how would a modeler be able to control the appearance of a mesh in tight corners if the renderer jumps over any tweaked gamma adjustments. Adding mesh resolution locally is not really a workable option since it also affects adjacent splines and we like the sparse models we can get with Hash patches.

Do you see my point? At which normal sampling resolution are we willing to live with? It is necessarily a compromise between the curves rendering we have now and the porcelain effect. Something in between. But where is this comfort zone? Too coarse and the mesh become difficult to tweak to get the complex curvature we want and then we have to add splines in order to increase the resolution or too fine and we have to deal with the creases again. :shrug:

Yves

jdates
02-08-2003, 07:21 AM
Say Joe, you sure do talk funny with all them big words!

Man, Joe.. do you dream in Binary!?

Well written.. and constructive too!
Makes me want to write a patch renderer now!

You rule Mr. Cosman.

jdates:thumbsup:

Kentaurus
02-08-2003, 09:52 AM
Originally posted by My Fault
OK, so I spoke to Steve today and here's the general gist of my conversation with him.
[...]
I still think there's room for additional modelling tools either from Hash or by the user community through the SDK to make modeling even better in AM.


In that case they will have to extend the scope of the SDK and probably also have to add some documentation...

I guess you didn't bring this up with Steve. Giving us a better SDK?

K.T.

My Fault
02-08-2003, 05:32 PM
Originally posted by Kentaurus
In that case they will have to extend the scope of the SDK and probably also have to add some documentation...

I guess you didn't bring this up with Steve. Giving us a better SDK?

K.T.

We did briefly touch on it. It sounds like it's on the list, but some of the other things like the renderer, stability, etc. are a bit higher right now.

I would think anyone who is working on or planning to work on plug-ins would already be talking to Hash. If not, you should probably do so.

dfaris
02-09-2003, 03:51 AM
What did he say they were going to do for all of us that have a broken app in v9.x?

PJC
02-09-2003, 05:13 AM
Originally posted by dfaris
What did he say they were going to do for all of us that have a broken app in v9.x?


Didn't Martin bring this up in his letter? Check that thread...

- PJC

marcopio
02-09-2003, 05:40 AM
Stability
Bug fixing and stability is at the top of their list, occupying around 70% of the programmers time right now. As for bugs he asks that if you do have any please email him directly and include any and all data. The chances of bugs getting squashed goes up exponentially when they have the project file to work with.

----

Well, I have submitted another but a few weeks ago and got a response from Hash Support, that they need exact steps to duplicate bug. Well I did give them exact steps to duplicate the bug. I gave them all of the project files that I kept having to recreate and still A:M was not working correctly.

The Projects I gave them, as soo as you import a Hash Model in the Choreagraphy, the object is listed on the properties but in the view you cannot see it. I am using version 9.51e

To this date, I have no response from Hash Support, not even to say they are still working on it.

See thread for the but and projects:
http://www.cgtalk.com/showthread.php?s=&threadid=39664

no0ne
02-09-2003, 09:32 AM
same thing happened about a year ago with me

sent bug-report
& answer was like : please give us some more info

sent in another mail with additional info
answer was like : please always send the whole bug-rep
& not additional info alone

ok, 3rd mail with link to a bug-rep-htm me set up
with ALL the steps & pics
& answer: no answer :wavey:

quite frustrated me started to work on something else
& half a year later, tada my name in the bug-fix-list & bug gone
guess i gotta work on my patience
but sad that i thought they'd not care
so me never reported another thing :surprised
til i saw they care for bugs but do not care to answer
i could live with that if it wouldn't de-motivate some bug-hunters

:beer:

Skevos_Mavros
02-09-2003, 03:41 PM
Actually, I usually end all my bug reports with the closing line "no reply necessary unless you need more information".

Interestingly, despite this Steve usually does send a quick thank you email anyway, though lately he hasn't replied. Maybe my last few bug reports came too thick and fast, or maybe they were too strongly worded (I also went through the whole "submit bug report three times before they saw it"). Or maybe they are really busy a the moment.

Or maybe he's just taking me at my word, and not replying unless he needs more info. As long as the bugs get fixed (one of mine was fixed in 10d) then I'm happy. If I want a pen pal, I'll look elsewhere! :)

jdates
02-10-2003, 04:34 AM
Originally posted by Skevos_Mavros
... As long as the bugs get fixed (one of mine was fixed in 10d) then I'm happy. If I want a pen pal, I'll look elsewhere! :)

I couldn't agree more! Great statement.

jdates


:applause: :applause: :beer: :applause: :applause:

HellBorn
02-11-2003, 09:32 AM
Nonproductive

Hellborn, your points are valid to an extent but again, I really think the Sub-D topic is dead.

You can read my other posts here - I am not, at present, a fan of Hash. However, the plain truth is that if you don't like splines - don't use A:M. It's really that simple. Splines are the foundation on which all of A:M is built and they are not going to go away...ever. Sub-D's are also not something A:M *has to have.* They would be very nice - since the animation tools are so sweet and many people would love to take advantage of those tools in conjuntion with their sub-d modeling skills. They are not needed though. You either accept splines in A:M or you use a dedicated app like Messiah or Kaydara Motion Builder with a package like Lightwave, Maya or C4D with poly models.


I did realise that AM is 'Hash splines only' long ago. That's why I'm leaving. It was OK before but today when even freeware and shareware applications can do SubD modeling the need for polygon support is needed at least if they want to increase the userbase. This bad polygon integration matter is one of the things I have to warn people about if they consider to buy AM. AMs feature list states that they can import polygons. Thats no lie because it can. It's just that what you get is more or less useless and a lot of rebuiling is needed. That they DON'T say.

Hmm..
If I use AM THEN I would be using a dedicated app and not the other way around as allmost all other applications can move models between each other.

I have checked some AM forums and actually 10%-20% of the questions concern polygon model import or export. So is a matter of importance.

If you checked my wish list you could see that this was not the top priority but I do find it more important than adding just another cool named feature to the list of features.

For Hash to get me back as an AM user and to stop warn people about AMs bad model integreation they have to either have full support for polygon models Or remove the creasing problem and a well working importer/exporter.


Here are two interesting quotes:

Nonproductive
I also disagree to a certain extent with your description of what an importer has to be. Because slpines are so sensitive to placement, bias, and direction I don't *want* an importer that makes those decisions for me.

JoeW
Splines let me "sketch" in 3d space. I don't think they require any more planning than subD's.

Ehhhh..don't now how comment on this except:
I don't dislike splines. It's just that the Hash splines are more or less useless compared to SubD becase of the creasing problem. If you get a edgeloop wrong or you want to increase density in a SubD model then thats no problem. Do the same in AM and there is a big chance you get a crease. OK then tweek that cp. But ohh no it was affected by the animation and arghh.... and you say no planning is needed....

If they fix the creasing problem AM will look a lot more attractive but I would still want to be able to use my polygon models and not rebuild everything. Life is simply to short for that.
I also think that because SubD applications cost little or nothing a lot of people will allready be confident in polygon modeling before getting an interest in animation. A lot of them wont be interested in an application that not supports what they allready have learned and not let's them use the models they allready have built.

In a year or two Wings might have animation. Maybe not as advanced as in AM but it would be free and might do excelent renders trough Rman or PovRay render engines.

If that happens, why should anyone buy AM?

That's what the people at Hash has to ask themself.

CGTalk Moderation
01-14-2006, 09:00 AM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.