KGtheway2B
![]() |
||||||
Posted: June 10, 2009 3:10 am | ||||||
KGtheway2B
![]() |
This is great!
I love when you can view the thumbnail and INSTANTLY know what something is. I'd suggest you add a switch somewhere that lets you render the image seamlessly without that nice looking gradient. It would be a lot more useful for 3d artists. Now: http://www.filterforge.com/filters/5423.html has a buddy! |
|||||
Posted: June 10, 2009 3:11 am | ||||||
Kraellin
![]() |
very nice
![]() i'm a bit surprised at the EP, since there are indications in the description that it doesnt seamlessly tile that well. but, congrats, nonetheless! ![]() If wishes were horses... there'd be a whole lot of horse crap to clean up!
Craig |
|||||
Posted: June 10, 2009 7:51 am | ||||||
Skybase
![]() |
Oh my... such awesomeness.
|
|||||
Posted: June 10, 2009 8:18 am | ||||||
Sign Guy
![]()
Posts: 554 |
Has anyone actually used this one yet? I ran a 3600 x 3600 render that took over 1.5 hours. Figured I was having a low on resources problem and rebooted my quad core. Now the second rendering is more than 2 hours into it and only half done.
Is this an incredibly slow filter or is something wrong with my system? Fred Weiss
Allied Computer Graphics, Inc. |
|||||
Posted: June 10, 2009 2:54 pm | ||||||
Indigo Ray
![]() |
It's not you. My time for only 600x600 is 8 min 50 sec! Besides the time, this filter is very resemblant of its name, nice job and welcome to FF! ![]() I'm not sure how you managed to make this un-tilable (the actual bumps, not the roundness), but it would be awesome if it was tilable. Keep it up! |
|||||
Posted: June 10, 2009 3:48 pm | ||||||
Sign Guy
![]()
Posts: 554 |
My one successful rendering tiles just fine. I'm running a second one now after stripping out the gradients and blends used to give it the roundness effect. After 22 minutes it is about 1/3 done. I would also point out that the default sampling on this filter is set to 25. I always set it to zero for the sizes I run.
I'll post back with the final time when the current render completes ... but at least it's progressing. Not sure if it's appropriate to upload the stripped down filter. I will if others feel it's acceptable. Fred Weiss
Allied Computer Graphics, Inc. |
|||||
Posted: June 10, 2009 4:41 pm | ||||||
KGtheway2B
![]() |
There's a sharpen followed directly by a blur, rip those out and it'll probably be much faster. There's also a few other blurs which could probably replaced adequately with a threshold, that and the gradient really should be an option.
Probably not ready for EP because of the seamless tile breaking gradient and poor construction. Fix those and It's got my recommendation. |
|||||
Posted: June 10, 2009 6:40 pm | ||||||
Targos
Posts: 196 |
I'll definately be making oranges if we can remove the gradient. Will be watching this for updates, or maybe more experienced forgers can mod it..
|
|||||
Posted: June 10, 2009 9:42 pm | ||||||
Moss |
Thanks for all the compliments guys. I guess nobody reads the description because then you would all realize that it can tile fine and you can remove the gradient with the Roundess slider. I also guess you didn't twiddle all the nobs. In the description I explain that the scaling method breaks tiling so IT ONLY TILES WHEN SCALE IS SET TO ZERO. I'm a noob to FF so that's the best I could figure out but from what I understand tiling is very easy to break in FF and it is hard to do scaling and rotating that doesn't break it. If you guys can suggest a better method for scaling I would be glad to implement it.
Actually when I was first going to implement scaling I was plugging the scal slider into all the scale settings for all the noise nodes, but then all the blurs and highpasses and other resolution sensitive things needed to be connected as well and I was doing all that but it was getting horrendously complex. For example some things needed to scale exponentially rather than linearily so I was having to plug the slider into a dummy checker node's colors and then use a gamma adjustment after that and then plug the result of all that into my blur radius or whatever. Anyway I was getting it all done but it was painstaking and the network was a mess and then I realized that performing the scaling towards the very end with offset nodes didn't cause the texture to repeat itself like I thought it would - it actually produced more random texture, so that was obviously a much better solution. Much later on I realized tiling didn't work when scaling was active but given FF's current limitations I think this is the best it can reasonably be? As for render speed. I rendered it at 2048x2048 with 5 samples in 4 minutes on a Pentium e2160 @ 3Ghz (dual core) with 4 or 3 GB of RAM, whatever Windows recognizes. I noticed it did use a freaking lot of RAM (1.5GB), so maybe that is the important thing to have. If you think there are more efficient ways of doing things I'd like to know. Like I said I'm a FF noob. I just found that I was running into a lot of limitations and had to do things more complicated then I would like. For example there is no way to adjust the density of dots with the Cell component so I had to stack a bunch of them together. My filter network grew exponentially larger as I tried to get it to look more and more realistic and as I added controls. I guess you have to pay some price for realism and flexibility. I hope this massive post has helped explain some things. Cheers. |
|||||
Posted: June 11, 2009 1:12 am | ||||||
Sphinx.
![]() |
Hi Moss,
This is indeed a great looking filter. Considered that this is your first submission, its understandable that you make a few construction mistakes (e.g. all those blurs, sharpens and highpasses). These bitmap based components break the true procedural flow and needs to allocate temporary memory blocks which is why you see that extreme memory usage. You should try to find some alternatives with curves. Furthermore you set the AA sample count to 25 - which is quite high. In the wiki you'll find a lot of articles on how to optimize filters and avoid/limit the use of bitmap based components. About the scaling issue: There is a general size/pixels slider that allows the user to scale down the general rendering, so as a rule of thumb, go for large features initially (or set the size slider lower before you start constructing) - then the user can scale down the features in the rendering if he/she wants to. I took a look at the filter and made some changes to show you a way to avoid the bitmap based components. The result is not identical, but that was not the goal either - I deliberately "over-optimized" this version to show where the main performance issues are. Feel free to take a look at the changes and use what you can from the optimizations. This version renders 600x600 in ~20 secs (vs 12 mins before). Citrus - Sphinx.ffxml |
|||||
Posted: June 11, 2009 2:36 am | ||||||
Sphinx.
![]() |
||||||
Posted: June 11, 2009 2:37 am | ||||||
Kraellin
![]() |
congrats on the EP, Moss and welcome to FF
![]() If wishes were horses... there'd be a whole lot of horse crap to clean up!
Craig |
|||||
Posted: June 11, 2009 8:20 am | ||||||
Moss |
Thanks Sphinx, that is much simplified and faster. The blurs were in there for a reason and I think mine looks noticeably better, but I didn't realize blurs were so resource intensive. Maybe adding just a bit of blur to your solution will make a happy medium. You guys must have slow computers because my render times compared to Sphinx's are 5 seconds and 27 seconds. Turning off the antialiasing on mine only drops it to 24 seconds.
That "size, pixels" slider does do the repeating thing that I was worried about, so it is not very useful unless you intentionally want something that repeats itself. I figured people deserve the ability to make more "zoomed out" textures not just extreme close-ups. With your simplified network with no blurs my original scaling method would be much more feasible. I should try it again. I didn't know what this "EP" was that you guys were talking about but I figured it out. Thanks, I'm honored. |
|||||
Posted: June 11, 2009 10:19 am | ||||||
Sign Guy
![]()
Posts: 554 |
FWIW, after experiencing render times measured in hours for a 3600 x 3600, I decided to try some smaller scale renderings for a comparison. This is on a quad core machine which i do renderings with on a daily basis. The results leave me wondering which version of Filter Forge you used to create this filter and if it could possibly be having some effect.
Here are the results using your second preset with roundness and scale set to zero: 600 x 600 = instantaneous 1800 x 1800 = less than one minute 3000 x 3000 = just over two minutes 3600 x 3600 = total choke, would not complete the rendering Since the less than Pro versions of Filter Forge have a 3000 x 3000 size limit, I am wondering if this might have something to do with the problems I am having. What version are you using? Fred Weiss
Allied Computer Graphics, Inc. |
|||||
Posted: June 11, 2009 10:50 am | ||||||
Sphinx.
![]() |
Yeah - definitely! It gives a nice subsurface scattering effect, a few bitmap based components won't make that much a difference either - specially if you ensure they're not in the same branch (i.e. one should not be a child of the other). Also you might be able to extract blur/highpass/sharpen from just one if the settings are close, check out this article.
I'm on the studio computer now (quadcore, 4gig): 3.2s (sphinx) vs 1.20m (original) vs 17s (original, no aa). Strange with those differences in our performance tests. If I change to a bigger rendering (than 600x600) things get really slow with the original though. |
|||||
Posted: June 11, 2009 10:56 am | ||||||
jffe |
----It flattens out a lot at a 0 setting, but it doesn't tile seamlessly. And really, it loses a lot of the illusion of being an orange/lemon/lime when ya flatten it out like that. Cool filter regardless, seems like one Crapa should've done, he's the local food-filter guy ha-ha. ![]() jffe Filter Forger |
|||||
Posted: June 11, 2009 1:38 pm | ||||||
Crapadilla
![]() |
Yummy!
![]() --- Crapadilla says: "Damn you, stupid redundant feature requests!" ;) |
|||||
Posted: June 11, 2009 2:39 pm | ||||||
Moss |
Woah, I just realized that my scaling thing does cause self-repetition just like the standard scaling. What the heck. I don't know how I missed that because I specifically looking for it. Guess I can take out that whole nonsense then and stop confusing people about the non-tiling.
So I guess making textures scale without repeating is no simple task unless you are using a simple network. Sign Guy, your render times sound comparable to mine. At the low resolution your quad cores are eating it up but at the higher resolutions RAM runs out and everything grinds to a halt. The highest I tried was 2048. I'm going to heavily revise this filter with a mix of Sphinx's simplified network. |
|||||
Posted: June 11, 2009 3:48 pm | ||||||
Indigo Ray
![]() |
It's OK, you can just say me. ![]() And I found the culprit to the tiling, or at least on Sphinx's. At the end, where the elevation gradient and the roundness are blended together, there is a gradient plugged into the opacity. Color 2 is the slider, and color 1 is a dark gray. Just set that to black and turn roundness down to 0 and it should tile. ![]() |
|||||
Posted: June 11, 2009 3:50 pm | ||||||
Kraellin
![]() |
moss, take a look at some of the zooming and shrinking snippets in the library. those will zoom and unzoom without repeats. feel free to add those snippets to your filter. just remember it's polite to give credit when you do. there is also a snippet for rotating.
If wishes were horses... there'd be a whole lot of horse crap to clean up!
Craig |
|||||
Posted: June 11, 2009 5:22 pm | ||||||
Sign Guy
![]()
Posts: 554 |
I cannot agree because I use 3600 x 3600 as a standard size for mastering and have rendered thousands of tiles at that size. Something is different with your Citrus filter that it would zip through at two minutes for 9 mega pixels and take hours at 13 mega pixels. Did you use the Pro version or one of the lesser versions? Fred Weiss
Allied Computer Graphics, Inc. |
|||||
Posted: June 11, 2009 7:09 pm | ||||||
Moss |
I'm using a demo of the pro version. I see what you are saying. You are saying that my filter is unusual compared to most in that it's render times scale exponentially?
I was just saying that your render times compared to mine make sense. |
|||||
Posted: June 11, 2009 11:53 pm | ||||||
Sign Guy
![]()
Posts: 554 |
Yes and no. There are lots of filters that have relatively slow render times, but they do not in my experience differ between a 9 mega pixel and a 13 mega pixel by so great a ratio. It would be more proportional. In the case of the sizes I tried, half again longer would be the normal case ... so at 3600 x 3600, your filter should render in about 3 minutes instead of 2. But what I am experiencing is greater than 2 hours. My concerns are not about your filter or workmanship. They are about what the cause of such an anomaly could be. It may relate to my quad core using XP Pro with 3.x GB of RAM being usable. So that perhaps I am hitting a threshold of overload that you are not. I would very much like to hear the render times of others using 3600 x 3600 and your filter with the roundness and scale set to zero. Then I would have a better idea if my system is at the heart of this. Fred Weiss
Allied Computer Graphics, Inc. |
|||||
Posted: June 12, 2009 12:46 am | ||||||
Sphinx.
![]() |
True, if you manually scale with offsets etc, you get a tile-repeating result. But the "size, pixels" slider is smarter than that: it doesn't repeat the rendering as if you scale down and repeat the texture - the variations of each "tile" are unique (that is: variations in components with a "variation" parameter). So actually the "size, pixels" slider does exactly what you want. The problem is that some non-variation components are tiled, which is mighty annoying. |
|||||
Posted: June 12, 2009 2:57 am | ||||||
Moss |
Ahhh, I see. So is there no way to prevent the roundess shading from repeating with the "size, pixels" slider? |
|||||
Posted: June 12, 2009 9:56 pm | ||||||
Moss |
I have updated the filter. It should run a fair bit faster, the network is way simpler now. I may have even made some visual improvements along the way.
I filled in the "Update Notes" section with some info and I gave some credit to Sphinx for helping me out but it doesn't show up on the webpage apparently. Is there a way to change the description without re-uploading? I swear I saw update notes on other people's filters... |
|||||
Posted: June 13, 2009 12:19 pm | ||||||
CFandM
![]() |
Great first filter Moss and congrats on that Ep to boot..
![]() Stupid things happen to computers for stupid reasons at stupid times! |
|||||
Posted: June 13, 2009 4:11 pm | ||||||
Kraellin
![]() |
moss, refresh your pages in i.e. just hit the refresh button here in the forums on these pages, and in the filter section where your updated filter is posted and you shld see the updates then.
If wishes were horses... there'd be a whole lot of horse crap to clean up!
Craig |
|||||
Posted: June 14, 2009 11:15 pm | ||||||
Moss |
Yeah, I see the updated filter. I just expected that the "Update Notes" would show up on the page, but I guess they don't.
|
|||||
Posted: June 15, 2009 12:13 am | ||||||
Kraellin
![]() |
ah, ok.
If wishes were horses... there'd be a whole lot of horse crap to clean up!
Craig |
|||||
Posted: June 15, 2009 12:55 pm | ||||||
SpaceRay
![]() |
||||||
Posted: August 10, 2012 7:31 am | ||||||
SpaceRay
![]() |
||||||
Posted: August 10, 2012 7:33 am | ||||||
Morgantao
![]() |
One of my favorite filters
![]() |
|||||
Posted: August 13, 2012 3:34 pm |
Filter Forge has a thriving, vibrant, knowledgeable user community. Feel free to join us and have fun!
33,711 Registered Users
+18 new in 30 days!
153,531 Posts
+39 new in 30 days!
15,347 Topics
+72 new in year!
18 unregistered users.