PDA

View Full Version : "bias normals" info from hash support


kungfudork
03-13-2003, 08:30 PM
hi all,
this came across the mailing list. i don't have access to the 10.5 alpha yet, so i can't commnet on it. i know some of you don't subscribe to the list , so i thought i would pass it along. i'm sure some of you have been testing this new feature.....i guess it is time to let hash know if it is a move in the right direction. anyway, i hope to join the ranks of 10.0 as soon as i can get my money save up!

"Bias Normals vs. Perpendicular Normals in V10.5a4
We had hoped to get some concrete feedback supporting this (radical) change in technology- We got some nice "oh this is cool" comments, but no concrete evidence that it was better, or that users preferred it with supporting data- If we don't get any input by Friday, we'll be going back to the traditional perpendicular normals rather than offering the bias normals option- Steve Sappington"

thanks,
kungfudork

Nonproductive
03-13-2003, 08:52 PM
This is quite possibly the most asinine thing I have ever heard from Hash. :rolleyes:

So let's see...if folks are not spending their time checking out a potentially fantastic new feature in an *early alpha* version of the software....you'll take it away forever?

The code has been written...it has been integrated into the render engine...the *work* is essentially done. However, because we all didn't throw a ticker tape parade for an *early alpha* feature - we're going to lose it.

Well...the one thing this *does* confirm is that Hash in fact expects users to use and embrace the *early alpha* versions of A:M - contrary to their earleir statements along the lines of "no one makes you use beta software."

Maybe they should wait until the software becomes a Beta instead of an *early alpha* version before making such rash decisions.

yeesh....

did I mention that this was an *early alpha*? just wanted to make sure...

koon69
03-13-2003, 10:09 PM
I see it that Hash wants to have this feature tested and have any issues ironed out. If no one is going to use it then they would concentrate on other issues. I think its a fair request. So many people complained - Hash listened and offered this - and no one is using it? Weird if you ask me. Im going to download 10.5 and see if I can show them something.

My Fault
03-13-2003, 10:31 PM
Yeah, but jeez give us more then a week to try it out. Many of us don't have the time to drop everything to try out new features.

More time Hash...more time.

Nonproductive
03-13-2003, 10:32 PM
Well...
I spent the last 9 years as a Project/Program manager and the short version of how I understand development is this:
Alpha: Implement
Beta: Test
Final: Release

I commend Hash for listening and implementing the new feature but you can not reasonably expect people to drop their "real" work to test a feature in an *early alpha* version. No one is using it because time is precious and dinknig around with an alpha feature is counter productive.

I would understand and be supportive if the statement was made during the Beta process...but not this early on.

This, in my opinion, comes across as "you all bitched, we did what you want, now use it or we're taking our ball and going home."

Roger Eberhart
03-13-2003, 10:49 PM
From what I can tell, the bias normals do nothing to fix the "Poissant" crease. If you make a "s" curve, as demonstrated on Yves page (http://www.micro-intel.com/users/ypoissant/tutorials/gammas.htm), then lathe it, you end up with nasty creases regardless of whether biased or perpendicular normals are used. So, if this is the problem that bias normals are supposed to fix, well, it failed.

Roger

koon69
03-13-2003, 11:12 PM
I agree that a week is no time at all. I plan to email asking for them to relax and just keep the feature. Im hoping the great folks here can email and tell them to keep it. I think in time this will fix the Poissant crease. But cmon give us more time and since you have the feature written just keep it.

daft
03-14-2003, 09:37 AM
I think this is the root of most problems with Hash. We as users are look at as Beta testers.
In my opinion Hash should hire some good artists that can challenge the program to be Beta testers.
I think most of other 3D programs developers have beta test and artist departments in their developing pipeline. I work with number of programs daily and it´s a fact that A:M is the most unstalbe software I have used. For example, I don´t remember when Max crashed last on me. A program that has much more options than A:M, I am not a programer but I would think that A:M would have a much lesser code than Max does. So don´t really see the arguments that some make in price diffrents between A:M and other programs when it comes to stability issues in A:M.

I don´t have the v10 so I won´t be able to try this new feature out. Sick of being a beta tester :)

Ps. just saw on cgchannel that Maxon has added HDRI support to Cinema. When do we get skylight and HDRI support in A:M?:shame:

amsmf
03-14-2003, 03:38 PM
>From: Steve Sappington <steve@hash.com>
>Reply-To: steve@hash.com
>To: animaster@animationmaster.com
>Subject: Bias (blurred) Normals
>Date: Thu, 13 Mar 2003 17:23:17 -0800
>
>
>Thanks to all of you who have provided useful feedback- for those of
>you working on providing actual project data to help us make our
>decision, we've extended the deadline to next Friday before we'll make
>our decision-
>
>Steve Sappington

So that's another 7 days to test.:shrug:

Wegg
03-14-2003, 04:28 PM
They are putting a deadlines on us provideing them with free "alpha" testing!?! Whats up with that?

hoochoochoochoo
03-14-2003, 06:31 PM
I am astounded at this approach. Even if there was truth that the "you bitched so you try the solution approach" - a deadline is not the way forward. Extended or not.

Creasing is a problem, whether you want to be a 1st or 2nd order modeller(ha ha ha) It exists, there are problems with 5pt patches or hooks on animation models. Too many users have complained in the past about creased models for anyone to say it's the modellers problem.

I had thought signs of progress where being made, I even started contemplating upgrading to this years subscription but this approach just means next year we'll still have someone (like me)complaining about creasing and how to solve it, Then they'll go to another bit of software which has relevant algorithms to help.
As I said on the "creasing thread" before being accused of starting an "everything is better than patches" thread, other software has ways of dealing with surface dis-continuity.
I for one will not bury my head in the sand and claim I use AM cos I'm an inherently superior human being.

AARGGGHHHH

balistic
03-14-2003, 09:09 PM
So, they're saying they're going to pull the . . . . feature if . . . *UNGH*. . . they don't . . . receive . . . *ARGH* . . . free immediate . . . alpha tes
http://www.bprince.com/HeadExplode.jpg

koon69
03-15-2003, 02:23 AM
I think its stupid to pull a cool feature that actually will help unless we test it for them. But to be fair theres what 4 people doing this at Hash? I figure lets send them some stuff and show them that we are listening - then ask them to make sure they are listening to us to. Fair?

Kevin Sanderson
03-15-2003, 01:52 PM
Well. like Yves said on the list, they have to find out if it's really working for people or not. Apparently it's a your mileage may vary kind of thing. Some people see a difference with their models, some don't. How can you do decent support for something like that? On the positive side, maybe extra input will help them find a working solution. As I've noted before, it seems like there's only a handful of people doing much, even with the alpha as you can gather from the limited number of names in the release bug reports and the limited posting of new work. And then there are all of us who don't have v10 yet and can't use the alpha or those who don't want to alpha or beta test, which is OK, too. So from that perspective, Hash Inc isn't getting a lot of good input.

Goon
03-15-2003, 05:59 PM
So does this mean that they have offered one solution given you two weeks to decide if you like it or not, and if you dont will say too bad, take it away forever, and work on another solution, or will they just say they tried and you have to live with it?

Wegg
03-15-2003, 06:10 PM
Sure sounds like that huh. . .

<sigh>

Kevin Sanderson
03-15-2003, 11:33 PM
Originally posted by Goon
So does this mean that they have offered one solution given you two weeks to decide if you like it or not, and if you dont will say too bad, take it away forever, and work on another solution, or will they just say they tried and you have to live with it?

I don't think it's that basic.

From what Yves said, and I would guess he's been communicating with more people than me, Hash Inc is trying to see what works.

Like Yves said, it's important to find whether it's a solution, a partial solution or not a solution at all that needs to be replaced by something better. But they won't know without data.

They need data, they need to see the creases, they need to see if this solution fixes those problems. It has for some, for others nothing...still creases.

If it doesn't work, let's move on.

My best guess is that Steve was probably trying to get some people off their butts to see if this fixed the problem or not. I gathered from the first message that they weren't getting much if any feedback and they have to move on or this version is never going to get out of alpha. After all, they had all this clamor about creases, they issue a fix and hardly anyone responds. Do you want them to blindly go on as if it does work and there are no issues? This is the opportunity to get these problems fixed NOW. They aren't going to get there by us just whining. Like I've said before and others have, too, they need to be pointed in the right direction. The problems have to be illustrated. We have to work within their paradigm for the time being...that means help from the users, not them hiring more people to test it in house. It's the reality of the situation. Waiting for a beta is even probably waiting too long.

If I had v10.5 alpha, and if I had a model that had creases, I would try it right now to see if it works or not. And then let Hash know. How hard is that?

koon69
03-15-2003, 11:42 PM
Excellent point. Ive sent them a simple model that showed some improvements. Im really glad thay are listening. Now that I have drank more tea - Im more relaxed! Damn that caffeine!

zandoria
03-20-2003, 06:48 PM
I did some tests of the biased vs. perpendicular normals, and sent this to Has support. I thought that I would post it here so that I could hear from other users if they agree with my conclusions an also for the benefit of A:M users who are not able to experiment themselves (the animation wasn't meant to be critiqued--just to to compare the difference between the surface interpretations)

I made a 10 second animation of my "Red Lori" model, without any hair or dynamic cloth, just walking in a circle in front of the camera. it is loopable if you want to play them side by side for comparison:
http://www.zandoria.com/images/Movies/perpendicular.mov is the first animation. there is no porcelain or biased normals--just the regular model.
http://www.zandoria.com/images/Movies/biased.mov is the second animation. It is identical to the first, but the skin group and the leather of the boots,gloves, and gorget have biased normals.

My thoughts looking at the renderings (besides needing some new smartskin at the shoulders), is that the flesh looks more organic in the biased version. If you compare the breasts, the biased version is much softer, and the connection to the torso is much smoother. The upper thigh is also more natural looking.

I am reminded of advice that I got when doing sculpture. I was told to make the clay look "full" as though it was expanding, this makes the sculpture look more "alive". I think that the rounded look of the biased normals helps recreate this feeling.

There are some 5 point patches at the area between the buttocks and lower back, and I think that the new averaged porcelain might be a good fix for that area, but I didn't experiment on this model with that. I did have good results with a combination of the bias and the porcelain with this picture of "bugs bunny":
http://www.zandoria.com/images/Bugs.jpg

I won't have any more time to experiment before Friday's deadline, but I like the control that this feature gives you over the way the surface is interpreted. I think that the "full" or "alive" look is much better for organic models, and I hope that you will find that it is worth keeping.

If you do keep this feature, my only request would be a toggle on the toolbar similar to the curved or peaked button for splines. That would make it easier to control how you apply this feature across a surface, without having to use named groups to keep yourself organized.

If you don't keep the biased normals feature, then I can always use named groups and porcelain to soften an area. But I know if that will replicate this look of "fullness", which I think (along with some good animation) can help bring a character to life. If the implementation of the biased normals makes the software less stable, then I would understand opting for only one type of surface interpretation. The other features that you have added such as multipass and averaged porcelain are great, so I'm happy.

William Sutton
Zandoria Studios
http://www.zandoria.com

PS. I have heard some people say that you should drop global illumination, but hey DON'T!!! :) Yesterday Justin emailed a link to a site called
www.boring3d.com and that guy had some great renders (of wierd shit), using global illumination, subsurface scattering, and really good depth of field. I have no idea what rendering plug-in he used, but the results were really good-- they might make a good benchmark to aim at.

Wegg
03-20-2003, 06:53 PM
Thanks for the visual test. It really wasn't as much of an improvement as I had hoped. :-/

zandoria
03-20-2003, 07:04 PM
I think that the ability to average the normals with a percentage of the porcelain material may have more promise. But for this test I didn't want to obscure the results of the biased normals alone...

I think a combination of the two techniques ( as in the bugs bunny) might be best.

My Fault
03-20-2003, 07:12 PM
The more I play with the bias normal feature, and see tests like you have done (great job BTW), the more apparent that this is not the "creasing" fix we were all hoping for.

One thing that kind of peaved me a bit in their description of the "blurred normals" on Hash's site was how they refer to "a popular and trendy “look” for computer-generated characters." So not having creases, which is a pathology of AM and no other software I've seen, is considered "trendy"? It almost seems like rather than admiting they have something to fix they are blaming this on the users. This seems like it would all be so much easier if Martin had some artistic tendencies and didn't approach everything like an engineer.

Wegg
03-20-2003, 07:18 PM
/me sees a little lightbulb go on over My Faults head. . .

balistic
03-20-2003, 07:21 PM
Antialiasing is just a trend too.

My Fault
03-20-2003, 07:28 PM
Originally posted by Wegg
/me sees a little lightbulb go on over My Faults head. . .

Haha, hey now, that lightbulb's been there for quite awhile Wegg, it's just been burning at about 5 watts until recently. I'm still not ready to sell my Chevy yet though :p

My Fault
03-20-2003, 07:31 PM
Originally posted by balistic
Antialiasing is just a trend too.

I don't understand how Hash can get one thing so incredibly right (animation/rigging) and one thing so utterly wrong (rendering). It's like they have two heads, one of which is a genius and the other is a total re-re.

Wegg
03-20-2003, 07:31 PM
/me waits till My Fault runs into the little black dots.

hoochoochoochoo
03-21-2003, 09:40 AM
well guys which is worse..... or better?

1/ the little black dots of death
2/ the untrendy creased model look
3/ the aliased render look

maybe we should have a poll
:D :D :D :D :D :D

[spo]
03-21-2003, 02:38 PM
"well guys which is worse..... or better?

1/ the little black dots of death
2/ the untrendy creased model look
3/ the aliased render look"


I'll let you know when it finishes finding patches.



:buttrock: :bounce: :applause:

KenH
03-22-2003, 12:50 AM
Perhaps there would be more eager testers if Hash hadn't been so unreceptive to critique in the past. If they are changing their strategy, they have to give notice and time.

My Fault
03-22-2003, 01:00 AM
I'm happy that they do seem to be making a real concerted effort to fix things. It may not be everything but it's a damn nice start. Kudos :thumbsup:

jayrtfm
03-22-2003, 02:19 AM
looks like the work in boring3d.com was done with vray
http://www.vrayrender.com/gallery/

Kevin Sanderson
03-22-2003, 11:19 AM
For those not on the list...this was in the message with bug fixes from Will Pickering

"10.5 Alpha Release 5

Further refinements are being made in our experimentation with surface smoothing controls.

Bias Normals

"biased normals" have changed, thanks to those customers who provided input. Try out the new implementation, make comparisons and give us your feedback.

Porcelain

Porcelain has been modified to increase the surface curvature but keep the smoothness. Please provide evidence that the "weight" control is useful because we are thinking of removing it since this latest change minimizes its influence.

Bias Handles

Many customers like A:M modeling "just the way it is", but have
suggested we make the Bias Handles manipulator more intuitive so they can tweak their models faster. So, now the bias handles can be dragged from any point-of-view, and Alpha, Gamma, and Magnitude change simultaneously. Also, either side of the magnitude can be adjusted separately if you press the <Shift> key. To adjust magnitude only (no angle), press the <Ctrl> key. The manipulator works in the old, one bias control at a time, method if you press the <Alt> key. Multiple keys can be pressed, such as <Shift><Ctrl> to adjust one side of magnitude only.

Watch For Errors

We have tried to make this 10.5 release as stable as 10.0. If you experience any slow-downs, rendering errors, or crashes that seem unfamiliar, please let us know."

Emphasis on the end is mine. Looks like they want to know of any problems now, not just repeatable ones for the alpha.

Kevin

zandoria
03-22-2003, 03:08 PM
They have made changes to the the alpha(5), which was posted to the list yesterday (Friday).

Several users are testing the surface smoothing features out, and hopefully this experiment will lead to better control for the rest of the community.
Here is an excerpt from the release note:

"10.5 Alpha Release 5

Further refinements are being made in our experimentation with surface
smoothing controls.

Bias Normals

“biased normals” have changed, thanks to those customers who provided
input. Try out the new implementation, make comparisons and give us
your feedback.

Porcelain

Porcelain has been modified to increase the surface curvature but keep
the smoothness. Please provide evidence that the “weight” control is
useful because we are thinking of removing it since this latest change
minimizes its influence.

Bias Handles

Many customers like A:M modeling “just the way it is”, but have
suggested we make the Bias Handles manipulator more intuitive so they
can tweak their models faster. So, now the bias handles can be dragged
from any point-of-view, and Alpha, Gamma, and Magnitude change
simultaneously. Also, either side of the magnitude can be adjusted
separately if you press the <Shift> key. To adjust magnitude only (no
angle), press the <Ctrl> key. The manipulator works in the old, one
bias control at a time, method if you press the <Alt> key. Multiple
keys can be pressed, such as <Shift><Ctrl> to adjust one side of
magnitude only."

Since Hash is working on the aspects of the program that everyone has asked for, try being patient...

[spo]
03-23-2003, 03:42 PM
"Since Hash is working on the aspects of the program that everyone has asked for, try being patient..."



if they repeat their usual pattern, they will be as concerned about user opinion just as long as they have to to appease the loudest.

they've already shown their lack of true respect by giving about 5 days to try out new features......"here..stop what you're doing and spend some time with our beta...or we'll PULL THE NEW FEATURES"

@$%# that.

not to mention you'll beta test them ....and end up having to pay for the version that has them working properly again. assuming they don't break everything else adding new features like they always do.


*patiently eats some worms*

zandoria
03-23-2003, 07:36 PM
They actually gave two weeks. And based on the feedback, they implemented some changes to the features--which I already posted about.

Maybe you could work on a project in whatever version of A:M you have, and then check back in a couple of months. Then the alpha and beta testing will be done in V10.5 and you will be able to benefit from it, and not have to resort to eating worms...(what is with that anyway?:shrug: )

HellBorn
03-24-2003, 10:39 AM
I don't think the 5 days where out of respect (based on some previus experience it could of course be so). I just think they had a new release coming (Alpha5) and wanted the input before releasing it.

Compared to before I actually prefere a Hash screeming for input than a Hash that dont seem to care :wip:

hoochoochoochoo
03-24-2003, 12:07 PM
hi hellborn - I sort of agree with you but I have a feeling the 5 days where a hangover of the old attitude to users.
I myself am still not convinced about upgrading to 10.5 but AM have at least listened to the creasing and surface anomaly problem.

Whether we all have the same opinion about the software guys or not there is small progress - even if they then make me blow my top occassionally afterwards.

HellBorn
03-24-2003, 12:09 PM
Well, your not alone;)

I'm also a bit in shock as I have now used AM 10 and 10.5 for about 15 hours without a single crash. I have to pinch myself all the time just to check that I'm not dreaming.

I also seems as we will get some working treatment against the creases.

So the real remaining worry is the renderer. It looks as they have a solution to the alias but as AM rather need a faster renderer than a slower the current solutions feels a bit dissapointing. I sure hope that they can find a way to cut renderspeed.

CGTalk Moderation
01-14-2006, 04:00 PM
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.