Daz 3D is part of 
Connect
DAZ Productions, Inc.
7533 S Center View Ct #4664
West Jordan, UT 84084
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.
Comments
Cool Spooky, thanks. I'll have to play around with them. What are the object accuracy settings to render the skin texture at full resolution at 15' away? If you can let me know that then I will work from there. If object accuracy fixes the problem at 720p high def then that is the perfect solution so we don't have to keep editting our textures.
Boojum the brown bunny
I have seen them in DS(3Delight).
At one time I use to use Carrara to create all my texture maps, and bake them out, for the models I give away, but found (sent images) that DS would render seams, but that depended on distance the object was from camera (they could appear/disappear with small movements of camera). I was not sure of the problem as the models/textures rendered correctly in carrara(7/8.1) and all other rendering applications I have. I did not want to install DS, so instead re-created the textures in another application which allowed me to place a margin (texture baked past UV map border). That resolved the issue. (Carrara, even with "Baker" plugin, does not give option for baking texture margins).
Personal opinion, I put it down to poor implementation.
DAZ_Spooky,
Thanks for commenting on this issue.
So far in my experience, and of course it is limited at this point, particularly with the Genesis and Genesis 2 textures I have available, the seam lines do not go away easily and certainly not with just a nudge. I have had far better results with many of the V4 textures (and I have no idea why – no surprise there) I have when applying them to a Genesis character. Many of those textures do NOT have any seams at all regardless of the distance the model is from the camera.
As far as complaints from customers go those types of things are subsets, of subsets, etc.
I know that I never saw a poll on this subject and I know I was never asked. In addition, IMHO, as more and more users become aware of this forced change, I believe more and more users will be upset over it. There are already some members attesting to that in this thread.
I remember this type of “reasoning” was used when then launched New Coke – how did that work out?
For my part I was totally unaware of issues with the textures being blurred and not suitable for close-ups. I have seen a great many close-up images posted on this forum (by PhilW and others) and did not see the blurring you speak of, nor did anyone complain of this issue along with the images they posted.
When these complaints were issued did the complainants say that they wanted a change to the default filtering method WITHOUT any way to choose another default method?
Did the complainants insist that they wanted Fast Mip Map without a way to easily pick and choose what object(s) it would be applied to?
Did the complainants want to have to spend hours and hours manually changing the Filtering on hundreds of shaders if they did NOT want Fast Mip Map?
I am not trying to argue for the sole use of one filtering method or another (or the merits and application thereof) but instead asking for an easy way for the end user to make the decision for themselves. Ideally the filtering method could be picked on a scene by scene basis and there would be the ability to easily pick a filtering method for different objects in a scene, so that one might have Fast Mip Map for objects in the distance and a different filter for close-up objects or any combination the user might prefer.
For my part I think it is important for the user to be able to make the choice and not have one method or the other forced upon them and not be penalized by a huge time suck if they do not want to use one method or another.
It should be all about choice and flexibility.
I found the texture map details, and embarrassingly enough they are documented - once you know where to look. Tweaking texture map settings will not be a problem at all.
The issue with the Object Accuracy setting, I have found that it has to be set to .5 pixels, is that it increases the render time dramatically.
If you look at the images in my first post here I mentioned that, in the case of the phone box, to get the same render results with Fast Mip Map that I get with Sampling (with the default Object Accuracy setting of 2 pixels) I have to set Object Accuracy to .5 pixels (for Fast Mip Map) and this more than tripled the render time.
I do not think I should have to experience increased render times to get the same result, with Fast Mip Map, that I can get by just using the Sampling filtering to begin with.
You are, once again, my hero Fenric.
I look forward to purchasing and using your solution.
You are the BEST!
Kakman and Brown Bunny,
I apologize. When I replied, I was completely skipping over the original complaint regarding the seems! Yikes!
In Spooky's defense, the beta has had the Fast Mip Map as default for as long as I can remember beta testing. Just none of us caught the seems problem. Please try and be civil toward him.
Kakman and Brown Bunny,
I apologize. When I replied, I was completely skipping over the original complaint regarding the seems! Yikes!
In Spooky's defense, the beta has had the Fast Mip Map as default for as long as I can remember beta testing. Just none of us caught the seems problem. Please try and be civil toward him.
I thought I was being civil.
If my reply to DAZ_Spooky came across as other than that, it was not my intention.
I am just trying to plead my case here.
I do feel that the implementation of the Fast Mip Map Filtering default (with no choice) was not well thought out, logical or well-reasoned - so maybe I was being over zealous.
I know that DAZ_Spooky is one of the good guys here and I meant no offense nor was I trying to attack him.
So to you and DAZ_Spooky and anyone else that feels I was not civil in my response to DAZ_Spooky – I sincerely apologize.
I was asked earlier to weigh in on this, so here are my thoughts. I think Fenric put it really well earlier that each of the filtering methods has its uses - for example, I just did a comparative test video with a detailed texture coming towards the camera. The Fast Mip Map version looked slightly blurrier at first, but showed much less flicker than the version with Sampled filtering, so I actually preferred the Fast Mip Map for animation. For a still image, I would probably have chosen the Sampling method in this instance. Closer to the camera, there is little or nothing to choose between the two. So my feeling is that one is not "better" than the other, they each have their uses.
So what we really need is a simple method to block change the sampling method - which hopefully is what Fenric will be able to produce. I would agree that I'm not 100% convinced that Fast Mip Map should be the default, but mainly because this is a major incompatibility with earlier versions of Carrara.
Okay, just doing a little more experimenting with the Fast Mip Map compared to Sampling.
As I understand it the “degradation” is supposed to “kick in” at a distance of some amount.
I made a quick scene just using the Old Beach Pavilion. I placed the building just far enough away from the camera so that the entire building is visible.
The top image uses Sampling. The bottom image uses Fast Mip Map.
I see much better detail in the top image, so I ask is this the way that Fast Mip Map is supposed to work? If so it just doesn’t make sense to me.
Please click on the images to see them in full size.
What about gaussian filtering? After I read Fenric's post I thought that maybe for those who want higher quality for closeup work and don't mind a bit of a hit on render times maybe gaussian would be the best option? I admit I also did a couple of closeup renders with very high render settings using Sampling, Mip Map, and Gaussian (and I turned off interpolation on each one too), but even flipping back and forth between each one I'm having trouble spotting any differences. Still I would like to know which would be the best for fine detail, if anyone can confirm. Oh and the render speeds weren't much different for any of the choices, honestly. Mip map was a little faster, but we're only talking a few seconds really.
Also does anyone know if antialiasing in Carrara is done by oversampling? I think that's usually the default method, but I don't want to assume.
I can't wait and I'm first in line to purchase :) Also eagerly awaiting to see your stuff come up in the PA sale.
And I'm dying to play with the new Skin Doctor, when it comes (hopefully very soon!).
I agree on all points! :)
Ok, I have done a bit more testing. Here are two renders with a bunch of V5's. All are using the default load of the Gia Elite skin. All are using default lighting for the scene. The first has an object accuracy of 2. The second picture has an object accuracy of .5. Each V5 character is 10' further from the camera than the one in front of it. So the distances are 13', 23', 33', 43', and 53'.
In the one with an object Accuracy of 2 the lines across the thighs are obvious for all characters. In the picture with the object accuracy of .5 the V5 who is 13' from the camera doesn't have the line, but all the others do.. the line across the thigh becomes more evident the further back from the camera you get.
The accuracy of 2 took 24 seconds to render. The object accuracy of .5 took 44 seconds to render.
Boojum the brown bunny
Thank you very much PhilW. As always your insights and opinions are greatly appreciated.
I can't wait and I'm first in line to purchase :) Also eagerly awaiting to see your stuff come up in the PA sale.
And I'm dying to play with the new Skin Doctor, when it comes (hopefully very soon!).
Well, I do not know how I lost my position in line - but why quibble?
I will be more than satisfied to be second in line.
Okay, another quick test. This time for a very SIMPLE close-up of V4.
Top image is with Sampling.
Bottom image is with Fast Mip Map.
I do not see much difference in these at all. I certainly do not see any evidence of the image (that used Sampling) being blurry.
In fact to me the “Sampling” image seems to have slightly more detail especially in the forehead area.
Again, I am just really trying to understand what the Fast Mip Map does and if it is working as intended.
In looking at the two V4 close-up images here in the forum (one below the other) it appears that the Fast Mip Map does render quite differently in regards to “skin tone”.
For ease of creation I created the scene in Carrara 8.1, saved it, and opened it in Carrara 8.5.
I then rendered the image in C8.5 – that is the top image (Sampling).
I then reapplied the same texture to V4 so as to have them load with the Fast Mip Map.
I then rendered the image in C8.5.
Overall, it appears to me that the Fast Mip Image is quite lighter. Even the grey color background appears slightly “brighter” in the Fast Mip Map image. It is almost as if the image with Fast Mip Map has a little gamma correction going on. Is that even possible?
Anyway it certainly seems that there is more going on with the Fast Mip Map than simple “distance degradation”.
Wow, that is a great representation of what is happening with the Fast Mip Map Filtering.
It makes me wonder even more if the Fast Mip Map is working as intended!
If you want to see the entire character in a scene, you really can’t be any closer than the character that is closest to the camera in your render and yet the “seams” are still very visible.
I know from some limited experimentation that not all textures are created equally.
I have several V4 textures that I have applied to Genesis in Carrara 8.5 (using a Poser Material Pose - .pz2, form the C8.5 browser) that have NO SEAMS whatsoever regardless of the distance that the character is from the camera.
Unfortunately, I am more perplexed then ever (not hard for that to happen- LOL).
Thank you so much for the time and effort you have expended regarding this “issue”.
Sorry that I missed your post initially, but I was posting to this thread at the same time you were.
ARRRRRRRRGGGGGGGGGG!!!! Ok, I have solved the problem... Evidently when you fast mip map the textures don't align perfectly anymore. I fixed the problem by making the area around the textures pink instead of white. I did this by using my eyedropper to select the color on the limbs near the edge, then flood filling the white. Ladies and gentlemen, I give you a render done with FastMipMap and the exact same settings as all the others I did, except this time the background of the Gabi texture maps used are pink instead of white, as is show on the attached thumbnails!!! And look, the seams are all gone!
Boojum the brown bunny
I went to the trouble of downloading both images so that I could flick between them - and concluded that there is no difference! (Not that I could notice anyway). Sometimes what you see is dependent on what is around it, and I think the effects you are seeing are actually an optical illusion - or you are just studying them for too long!
As Fast Mip Map produces lower resolution versions of textures to use when the texture is displayed smaller, I would not expect to see any difference when the texture is close to the camera.
Thanks for that, Boojum - good to have a fix, but that's going to be a lot of work if I had to do that with all my textures!! OK, only when needed, I know.
Agreed. I did not notice a difference when I went back and flicked between the images using my photo editor.
Thanks for the information and the explanation.
Kakman and Brown Bunny,
I apologize. When I replied, I was completely skipping over the original complaint regarding the seems! Yikes!
In Spooky's defense, the beta has had the Fast Mip Map as default for as long as I can remember beta testing. Just none of us caught the seems problem. Please try and be civil toward him.I caught it when Fast Mip-Map was introduced. :) Which is how i know that answer. The Devs and I had some long discussions about it. :) After all I am the Software QA Manager, not a dev. LOL. I have to look at software from a different perspective than they do.
There was no poll, it was a large stack of requests and reports in the bug tracker and we still went back and forth on it.
In most cases it produces faster and better results, note i said most cases. The default is the "80%" case and improvements were made in to make it work better than when we first sat down with it.
You guys, who clearly know what you are doing, already know how to change it are more than welcome to change it. We expect you to. If you are modeling your own stuff, we figure you will set your shaders for the look you want anyway. For the person less experienced, this setting is believed to give a better, out of the box, experience in a wider number of cases.
Hi Spooky..
Some objects, like the Genesis Super Suit, have hundreds of texture maps. There needs to be an easier way to set this setting than hand changing all 100 to not be FastMipMap. That said, It looks like the setting the background of the skin texture to roughly the same color as the main map fixes the problem, so I'm less worried about it now.
Boojum the brown bunny
Speaking of which, the Shader system in Carrara really should use the Material Zone name, instead of a straight index. We have been discussing fixing that, however, fixing that will break current multi-shader presets. The upside is that, for example, you could apply the same preset designed for V4 or M4, etc. to Genesis. It will also fix applying a preset to V4 regardless of whether V4 was loaded as part of a scene in a DUF or loaded directly as a CR2.
Comments on this?
While I don't completely understand the ramifications of what this change in the Shader system would have, I believe that it's important to maintain backwards compatibility with the vast library of Shader presets that are currently in use. Perhaps multi-shader presets could be automatically converted to prevent "breakage".
While I don't completely understand the ramifications of what this change in the Shader system would have, I believe that it's important to maintain backwards compatibility with the vast library of Shader presets that are currently in use. Perhaps multi-shader presets could be automatically converted to prevent "breakage".The problem is there is no information in the shaders that tell you what material zone they are supposed to apply to. It is strictly an index, where the first Material Zone on the object gets the first sub-shader, etc..
So if the First Material zone, when you saved it was teeth, but the first material zone when it was applied the next time was face....
BTW there is a reason you don't have this fix with 8.5 and that is because we haven't figured a way to avoid the incompatibility.
True, but that is if you want to change a selection of them. What I suggest you do is have a way to change the filter settings on ALL the maps on an object. There is a very good reason for this....
If your wanting to render an object without loss of quality due to FastMipMap... then your almost always wanting to set all the textures on the object to not use fastmipmap. That includes the eyes, the lips, the fingernails, and everything else. You don't normally want one part of a model to be high quality and the rest to be less quality. Also, this gives a person a choice. If less than half the textures need to be changed, then they can do it individually. If more than have need to be changed then they can change them all and change the ones they don't want to use a different filter back.
But honestly, I suspect everyone is going to want to be able to set all to a particular filter setting.
Boojum the brown bunny