chaos_crystal
|
|
« on: December 31, 2012, 01:25:12 PM » |
|
I have my first halfway decent image created in Apo 7X and left it at 1024 x 768 to show detail. The density/quality setting was formerly, 4000, but the rendering of said image still looked bad, all pixelated. Other settings: Filter radius at 1, Oversample 4. I have tried a smaller test image with FR at 1.5 and Oversample at 5, but it didn't seem to make much diff. I thought I would try another rendering, still at 'wall size' dimensions, but with the density setting at 100,000. That would have eaten up 192 MB memory and taken all of 42 hours. And my puter was slowing down so badly I couldn't have used it for anything else. Is there any hope for me to get really good images out of Apophysis? This is nuts!
|
|
|
Logged
|
I conceived and developed a new geometry of nature and implemented its use in a number of diverse fields. It describes many of the irregular and fragmented patterns around us, and leads to full-fledged theories, by identifying a family of shapes I call fractals.
— Benoit Mandelbrot
|
|
|
David Makin
|
|
« Reply #1 on: January 02, 2013, 12:10:40 PM » |
|
I have my first halfway decent image created in Apo 7X and left it at 1024 x 768 to show detail. The density/quality setting was formerly, 4000, but the rendering of said image still looked bad, all pixelated. Other settings: Filter radius at 1, Oversample 4. I have tried a smaller test image with FR at 1.5 and Oversample at 5, but it didn't seem to make much diff. I thought I would try another rendering, still at 'wall size' dimensions, but with the density setting at 100,000. That would have eaten up 192 MB memory and taken all of 42 hours. And my puter was slowing down so badly I couldn't have used it for anything else. Is there any hope for me to get really good images out of Apophysis? This is nuts! Some flames will take much, much longer than that to get decent quality everywhere. Flames are generally 10 to 100 times slower to generate than standard escape-time fractals so unless you have a really fast CPU (e.g. dual 6-core or better) then you'll have to put up with long render times for high res. in good quality. I'd expect a flame fractal say at 4000*3000 pixels to take up to a week (non-stop rendering) on say a core2duo - longer if using a lot of modifiers, even up to a month. My original render of an escape-time fractal "Aquamariine" at 8000*5000 pixels took 29 days - though granted that was in 2003
|
|
|
Logged
|
|
|
|
chaos_crystal
|
|
« Reply #2 on: January 03, 2013, 12:00:07 AM » |
|
I MAY have stumbled onto at least a partial answer to the problem. And mind you I'm not expecting to render anything any larger than maybe 1024 x 1024 or so...I don't think my old 32 bit could take it w/o collapsing... Really, I feel like I'm in a time warp, this puter would probably have been 'standard' in 2003 but here we are in 2013 and I'm still struggling on this old dinosaur. Anyway...somewhere I saw mention of leaving the 'transparency' option unchecked unless you want a white background that can be used in alpha channel in graphic arts. Must have been on a video tut. But it didn't sink in till later...when I realized that all my pitiful efforts thus far have, sure enough, the white background. BUT the resolution looks okay! The light went on, or at least began to flicker in my brain... This could be my main problem. If an image has the transparency option ticked before rendering, then apparently it is rendering for transparency needs....duh!!...but NOT for a black background. Maybe that's why no matter what I do, those bits and bobs are scattered everywhere when I look at my images on the black background. I'm about ready to try another render (even if, granted, it takes many hours *sigh*) with transparency UN-ticked. Some people just take longer to "get it" than others do.....lolol
|
|
|
Logged
|
I conceived and developed a new geometry of nature and implemented its use in a number of diverse fields. It describes many of the irregular and fragmented patterns around us, and leads to full-fledged theories, by identifying a family of shapes I call fractals.
— Benoit Mandelbrot
|
|
|
Dinkydau
|
|
« Reply #3 on: February 09, 2013, 04:39:05 AM » |
|
Your filter radius is too high. Try 0.4. Personally I use 0.2, but the lower the filter radius, the more points you will need to calculate. 0.4 should be enough for sharp images.
|
|
|
Logged
|
|
|
|
chaos_crystal
|
|
« Reply #4 on: February 15, 2013, 06:35:25 PM » |
|
I'm glad I caught this note...I think most recently I've been using 1 for filter radius. Need to get back to practicing with Apo anyway...MB 3D render times are killing me AND my old computer!...we're both getting way too old for that, LOL. So I will lower the filter radius a little, and see what we get!
|
|
|
Logged
|
I conceived and developed a new geometry of nature and implemented its use in a number of diverse fields. It describes many of the irregular and fragmented patterns around us, and leads to full-fledged theories, by identifying a family of shapes I call fractals.
— Benoit Mandelbrot
|
|
|
lycium
|
|
« Reply #5 on: February 15, 2013, 07:33:04 PM » |
|
Please consider giving Chaotica a try, one of its main goals is to deliver the highest possible image quality: http://www.chaoticafractals.com/forum/viewtopic.php?f=6&t=42Besides this it also renders a lot faster than Apopohysis/flam3, however it doesn't support the 3D hack and DC stuff (these kinda things will be implemented properly in the native editor). Some other things maybe worth noting: 1. Apophysis' multithreading is quite broken/incorrect, and though I've pointed out some bugs to the authors it doesn't seem like it'll get fixed. You get very different results rendering with 1 thread, 1 thread in a VM with only 1 thread (!!), and rendering with 2+ threads. 2. Apo's filter radius seems to be resolution-dependent, i.e. one setting won't work for all resolutions. 3. You can't adjust any of the imaging stuff while it's rendering, need to restart from scratch. Etc...
|
|
« Last Edit: February 15, 2013, 07:35:43 PM by lycium »
|
Logged
|
|
|
|
lycium
|
|
« Reply #6 on: February 15, 2013, 07:49:22 PM » |
|
Flames are generally 10 to 100 times slower to generate than standard escape-time fractals [citation needed] (And of course, this is a joke, since a statement as woefully unqualified as that is pretty much like saying "Audis are 100x faster than Volkswagens" or something XD)
|
|
|
Logged
|
|
|
|
Apophyster
|
|
« Reply #7 on: February 16, 2013, 04:40:06 AM » |
|
1. Apophysis' multithreading is quite broken/incorrect, and though I've pointed out some bugs to the authors it doesn't seem like it'll get fixed. You get very different results rendering with 1 thread, 1 thread in a VM with only 1 thread (!!), and rendering with 2+ threads.
2. Apo's filter radius seems to be resolution-dependent, i.e. one setting won't work for all resolutions.
Are there examples to support what you say? Are there side-by-side comparisons of Apo vs. Chaotica renderings, including parameters? Are you referring to *all* versions of Apophysis in your remarks, or is there a particular version which exhibits the faults you list?
|
|
|
Logged
|
Rice, wheat and corn make the world go round. Love and money are just passengers. Friendliness is the destination.
|
|
|
chaos_crystal
|
|
« Reply #8 on: February 16, 2013, 05:37:36 AM » |
|
Please consider giving Chaotica a try, one of its main goals is to deliver the highest possible image qualityI certainly will try Chaotica, lycium! My quest for high quality fractals with a minimum of struggle seems endless...but so is my fascination...
|
|
|
Logged
|
I conceived and developed a new geometry of nature and implemented its use in a number of diverse fields. It describes many of the irregular and fragmented patterns around us, and leads to full-fledged theories, by identifying a family of shapes I call fractals.
— Benoit Mandelbrot
|
|
|
lycium
|
|
« Reply #9 on: February 26, 2013, 10:01:15 PM » |
|
Are there examples to support what you say? Are there side-by-side comparisons of Apo vs. Chaotica renderings, including parameters? Are you referring to *all* versions of Apophysis in your remarks, or is there a particular version which exhibits the faults you list?
Hiya, sorry for the delay in answering these (good, directed) questions! I'm working on a detailed answer including parameters and renders for these problems, among others. Regarding the filter radius problem, simply leave the parameters and render them twice at a different resolution; the difference in filtering "softness" should be readily apparent.
|
|
|
Logged
|
|
|
|
Apophyster
|
|
« Reply #10 on: February 27, 2013, 12:43:59 AM » |
|
Oh cool! Examples always help to support claims!
It could be that my eyes are too old to see a "softness" difference. Probably also you know, I'm mostly an anim fanatic, and don't do many large renderings. I couldn't get Chaotica, the latest I think, to run. Now here's my bad: I don't have an example of the error message to show you. If I can remember, and I can squeeze in a few moments, I'll let you know more details at DA.
I also apologize for any awkward or harsh words or expressions I may have used formerly when trying to communicate with you (or anyone else for that matter). Life's been a beast lately...
Fred
Rice, Wheat and Corn make the world go 'round. Money and Love are just passengers. The destination is friendly relations among all.
|
|
|
Logged
|
Rice, wheat and corn make the world go round. Love and money are just passengers. Friendliness is the destination.
|
|
|
toxic-dwarf
|
|
« Reply #11 on: March 31, 2013, 01:52:39 AM » |
|
Anyway...somewhere I saw mention of leaving the 'transparency' option unchecked unless you want a white background that can be used in alpha channel in graphic arts.
Yup took me while to figure that one out. I was so frustrated that my previews looked good, but the render looked awful. Then i disabled transparency and the render looked spot on. To be fair I quite often now render with transparency enabled so I can do some nice post work in Photoshop, you can just paint a background layer and stick the render on top and it looks just like it should. Also I read that it's a good idea to tick the 'Postprocess render' selection in the 'render flame to disk' window. That way once it has finished the main render you can adjust filter radius, brightness, gamma etc before saving. Here's a link to a nice tutorial on render optimization - http://www.ultragnosis.com/fractals/Resources/rendopt.pdf
|
|
|
Logged
|
|
|
|
chaos_crystal
|
|
« Reply #12 on: April 04, 2013, 12:32:55 AM » |
|
Yup took me while to figure that one out. I was so frustrated that my previews looked good, but the render looked awful. Then i disabled transparency and the render looked spot on. To be fair I quite often now render with transparency enabled so I can do some nice post work in Photoshop, you can just paint a background layer and stick the render on top and it looks just like it should. Also I read that it's a good idea to tick the 'Postprocess render' selection in the 'render flame to disk' window. That way once it has finished the main render you can adjust filter radius, brightness, gamma etc before saving. Here's a link to a nice tutorial on render optimization - http://www.ultragnosis.com/fractals/Resources/rendopt.pdfKEWL!! As I recall, I have transparency enabled too, and if it will help to place the finished image on a background gradient or some such, I'm all for it! And yeah, I like post-processing and tweaking around, that should also help. And thanks a ton for the link. There has gotta be a means of putting even those random images to good use, those we all hit upon sometimes that may be random but still look good enough to use... Rendering isn't THAT difficult, no reason for it to cause unnecessary stress...
|
|
|
Logged
|
I conceived and developed a new geometry of nature and implemented its use in a number of diverse fields. It describes many of the irregular and fragmented patterns around us, and leads to full-fledged theories, by identifying a family of shapes I call fractals.
— Benoit Mandelbrot
|
|
|
|