Jump to content

Wikipedia:Featured picture candidates/POV-ray

From Wikipedia, the free encyclopedia
Image generated programmatically with aid of a computer.

As per discussion below, I nominate this awesome, completely "synthetic" image. It looks so natural, with all reflections and refractions, that you really have to look hard to see it's artificial. It's in POV-ray. (And, it's actually excellent as a still life, too!)

NOTE re closing this nomination: We are still waiting for two new renderings. Should we keep this here long enough to include them?

Hey, many real life objects can be completely made up of primitive shapes combined via CSG, like dices indeed. There are dices without round corners, you know? And i don't see the pouring thing you're talking about. A truly photorealistic work, if you ask me... (namekuseijin (at) gmail.com) — Preceding unsigned comment added by 201.40.20.130 (talkcontribs)
I have asked the artist, User:Gilles_Tran, who created this image, to reply here. --Janke | Talk 06:31, 17 March 2006 (UTC)[reply]
Thanks for the nomination (though people looking for a real still life master working with POV-Ray should have a look at Jaime Vives Piqueres' work)! The picture took a little less than 2 days to render at 1024*768 (blame the focal blur). I've put a resized version on line to smooth out some of the graininess in focal blur and some of the poor antialiasing, but the original version is available here. Of course, folks with better hardware than mine (P4 3Gz) can also re-render it at a larger size. When POV-Ray 3.7 is out (with true multiprocessor support), it will be even possible to give the scene code to a render farm and render a giant version of it. --Gilles Tran 10:03, 17 March 2006 (UTC)[reply]
I have access to two "supercomputers" (32 Xeon 2.4 GHz CPUs inside) that run Linux. Is POV-Ray going to get a multiple-processor version for Linux soon? --Dante Alighieri | Talk 17:48, 17 March 2006 (UTC)[reply]
Right now the Windows version serves as a platform for the beta version and there's no Linux and Mac 3.7 beta available. There has been massive changes to the core of the program (to add multiple-processor support among other things) so debbugging it gets top priority. Once the core is out of beta, the POV-Team should roll out the OS-specific versions (and the source code), including of course the Linux one. No deadline set though. Note that there are ways to run POV-Ray scenes on several machines, but there has been some issues until now with scenes using radiosity like the "glasses" one. 3.7 should solve that. --Gilles Tran 19:49, 17 March 2006 (UTC)[reply]
I decided to try and render the image at 3200x2400 on my Athlon64 3000+ (1.8GHz). It should be done in 2 days a long time. ~MDD4696 23:22, 17 March 2006 (UTC)[reply]
We'll be waiting for the upload! --Janke | Talk 09:02, 18 March 2006 (UTC)[reply]
Nice! Be sure to use an appropriate assumed_gamma value otherwise the image could be too pale. The scene uses assumed_gamma = 1 because I'm working on LCDs with a display_gamma = 1 in the POVRAY.INI file. --Gilles Tran 09:31, 18 March 2006 (UTC)[reply]
Aw, shucks. I had been rendering it without a Display_Gamma in the php.ini and with assumed_gamma 1. It looked fine on the rendering system, but was too light on my Windows laptop (which I know has a gamma of 1). The Linux system I am rendering it on has a really crappy monitor and video card (PCI!), so using gamma.gif as a reference I would say it has a gamma of 3.2. I have now set Display_Gamma in php.ini to that. I tried reading the documentation, but I'm still not sure what to set display_gamma to. Should I leave it at 1? For comparison. ~MDD4696 17:36, 18 March 2006 (UTC)[reply]
Try to have assumed_gamma = display_gamma and see what happens (the doc says "POV-Ray allows you to specify in the scene file the display gamma of the system that the scene was created on"). Not that I'm sure that is going to work... The gamma problem in POV-Ray has been a sore point and is currently being addressed in the next version to make it more manageable. --Gilles Tran 18:09, 18 March 2006 (UTC)[reply]
In general, computer monitors have a gamma of 2.5 or so (Macs are 1.8), so maybe that would be the right value? You can test your own monitor with the scale in the gamma correction article. --Janke | Talk 07:19, 19 March 2006 (UTC)[reply]
Ok, things are chugging along fine now. The first 252 lines (out of 2400) took 10:26:50 to render. ~MDD4696 06:16, 19 March 2006 (UTC)[reply]
  • Not release date set per official policy, as tentative release dates have been proven unreliable and problematic in the past (real life keeps getting in the way). The current beta expires on April 1st so we may have a better view then --Gilles Tran 13:16, 20 March 2006 (UTC)[reply]
It's not the index of refraction that gives it away; for real ice it's 1.31, and in the render it's 1.33 (the same as water). I think the problem is that real ice cubes are very imperfect; they have cracks, parts that scatter a lot of light, etc. But I agree — despite the ice the image is very convincing! —Deadcode 19:47, 26 March 2006 (UTC)[reply]
I think we can assume amazing means support... ;-) --Janke | Talk 07:34, 19 March 2006 (UTC)[reply]

Discussion closed as of 06:14, 31 March 2006 (UTC). Additional comments may be made below.


  • sorry, man. bad luck: it's way too dark. start over... ( namekuseijin (at) gmail.com ) —The preceding unsigned comment was added by 201.40.20.130 (talkcontribs) 16:59, 31 March 2006., referring to MDD's rendering
  • I can no longer say that no highlights are blown... the highlights in the stems of the glasses are extremely bright. —Deadcode 19:10, 2 April 2006 (UTC)[reply]
    • Interesting... my render has now passed MDD's percentage-wise, even though I'm only running a 32-bit binary on my Athlon64 4000+. MDD, you have antialiasing turned on, don't you? I have it turned off. The only difference I notice is that your render has a smoother highlight along the left side of the cone of the right wine glass. I have to admit that your higher resolution makes up for the graininess of the focal blur; I should've gone your route instead of increasing blur_samples. —Deadcode 22:30, 4 April 2006 (UTC)[reply]
      • I really don't know much about POV-ray, so I couldn't tell you if I had antialiasing on. Reassuring, huh? I installed it, and then hit render. The only thing I tweaked was the assumed_gamma. I saw your note about posterizing in the history view. You're right--anything above a minor gamma boost produces significant posterizing. Would I have been able to render the image with a higher-bit color, had I known? ~MDD4696 02:02, 10 April 2006 (UTC)[reply]
        • I just saw that you rendered the image with 3.5 and this is what causes the differences in highlights/brightness: the clipping of bright (> rgb 1 white) radiosity data was removed in 3.6, making possible HDR-like images with typical "burnt" highlights (caused by a bright sky for instance). Note that when focal blur is used antialiasing has no effect so it's unecessary to turn it on (for all versions including 3.6).--Gilles Tran 09:58, 10 April 2006 (UTC)[reply]
    • As of 2006-04-11 21:59 (UTC), my 2048×1536 render has finished. It took 560 hours. (Note that I changed the ice cube's index of refraction to 1.31 like real ice, but this doesn't make it look any more realistic.)
Here is a version matching Gilles Tran's render in luminosity, with its linear gamma and strong clipping of highlights
Here's one with very conservative clipping of highlights, lightened for monitors that have a gamma anywhere near 2.3 or so
Here is the raw linear render in 48-bit color, at 1/4 the brightness of Gilles Tran's render to reduce clipping of highlights
Suggestions on what kind of version to finalize would be welcome. —Deadcode 22:59, 11 April 2006 (UTC)[reply]
Great to see the finished pic! Personally, I'd go for the original version with the stronger highlights since this particular effect (due to the non-clipped radiosity values new in 3.6) was part of the demonstration and it will be less confusing for users if the image matches the available scene code. In fact it's often the rule in certain 3D circles that scenes must not be post-processed, so that viewers can appreciate fully the abilities of both the rendering engine and the 3D artist, otherwise it's a demonstration of image editing skills rather than 3D ones. Of course, this "rule" doesn't make sense when the image is meant to be appreciated independently of its technical origin (for commercial or artistic reasons, particularly). In any case, if people prefer the clipped version from an aesthetical point of view, I don't see that as a problem as long as a link to the straight-from the-renderer, unprocessed version is also provided.
The problem with the original version is that it has a gamma of 1.0, and most people use their computer at a gamma of at least 2.2. The only postprocessing I did was to apply a tone curve and boost color saturation; surely that doesn't count? It's nothing more than digital cameras do, even professional ones. —Deadcode 17:44, 13 April 2006 (UTC)[reply]
About the ice cube, it can be made more realistic by adding some surface normal perturbation. I'm testing that right now (on the ice cube zone only of course), but the render times go through the roof due to the extra calculation so it won't be finished until tomorrow. The ice cube render zone is +sc0.275 +sr0.665 +ec0.395 +er0.82 and the additional code I'm testing is normal{agate 0.25 scale 0.5} (to be inserted in the T_Ice texture definition). BTW the fact that testing glass scenes takes ages is also one of the reasons why it's relatively low in detail... --Gilles Tran 13:49, 12 April 2006 (UTC)[reply]
  • Here's a new version with a (possibly) better-looking ice cube as described above. I just rendered the ice cube (9 hours...) and pasted the rendered part on Deadcode's version.--Gilles Tran 14:01, 13 April 2006 (UTC)[reply]
    • Nice paste job, but I don't think that looks any more like real ice. I think what is needed is to give it an unevenly frosted interior (subsurface light scattering). Can POV-ray do that? (BTW, what's the best way to render a crop like you did, such that the pixels correspond?) —Deadcode 17:44, 13 April 2006 (UTC)[reply]
      • The media{} feature in POV-Ray is meant to give an interior to objects (these icecubes use media (and photon mapping aka caustics)). The drawback of media{} is that it's extremely slow to render and difficult to tweak, so that tests alone are likely to take hours. Unfortunately, the features that would make the ice cubes more realistic (scattering media, blurred reflections, photon mapping) are also among the most render-intensive, and using them in a scene featuring glass, radiosity and focal blur is really asking for trouble...
To render a partial image, use the +sc/+sr/+ec/+er values given above (+sc = start column, +er = end row ; the value is given either in pixel or in percentage/100). In Windows, the partial output coordinates can be set automatically by shift-drawing a rectangle in the render window.--Gilles Tran 19:30, 13 April 2006 (UTC)[reply]
  • Support. That is a stunning picture, if I didn't know in the first place, I never would have guessed it was a rendering. Nhandler 06:40, 7 April 2006 (UTC)[reply]
  • Oppose. Looks like nice ray tracing but not very interesting. No obvious use of more interesting computer techniques. Not much detail. Needlessly imitates small camera with limited depth of field. David R. Ingham 06:53, 7 April 2006 (UTC)[reply]
    • More interesting techniques such as? Buddy13 00:11, 8 April 2006 (UTC)[reply]
    • Small cameras have lots of depth of field. It's the large cameras that allow you to have a shallow depth of field (or more precisely, cameras with large sensors and big lenses). Of course, they don't force it on you — the lens's iris can be closed down for a wide depth of field. But close it down too far, and the picture actually gets blurrier due to diffraction (POV-ray doesn't simulate this, incidentally) which means in reality, you can't have everything in perfect focus at once. So with no focal blur at all, the render would look unrealistic. At issue may be the amount of blur, but IMO it makes this render look like a photo taken by a professional camera. What I would criticize is that POV-ray forces the aperture to be square-shaped, whereas it should be circular. —Deadcode 18:51, 8 April 2006 (UTC)[reply]
    • A couple of remarks about David's criticism: the picture was conceived in 2004 as a demonstration scene for POV-Ray users that would feature HDR-style radiosity (but without a HDR bitmap) and focal blur. It has an educational - rather than artistic - purpose and is voluntarily limited in features and detail so that POV-Ray/Wikipedia users can test and modify the scene easily (in fact, the objects were originally created for this scene). It is not meant to be a demonstration of state-of-the-art CG, something that would require more powerful (and expensive) tools, and should not be presented as such if the image ends up as a featured picture. It is just an image that anyone can create using a free raytracer.--Gilles Tran 14:21, 10 April 2006 (UTC)[reply]

Vehemently and fiercely oppose, delist even - I dislike this image intensely, it's an awful demonstration of POV-Ray. Seriously though, support. Even though the other two have a 6-day headstart, I started rendering a 1280 x 1024 version of the image for a desktop background about 12 hours ago. 15% complete. ;) Nippoo 10:05, 11 April 2006 (UTC)[reply]

Poll

[edit]

Since I originally nominated this, I feel a responsibility to get the nomination finished, too.

We now have five versions to choose from:

  1. Here is a version matching Gilles Tran's render in luminosity, with its linear gamma and strong clipping of highlights
  2. Here's one with very conservative clipping of highlights, lightened for monitors that have a gamma anywhere near 2.3 or so
  3. Here is the raw linear render in 48-bit color, at 1/4 the brightness of Gilles Tran's render to reduce clipping of highlights
  4. new ice cube version
  5. An edited version of #2: slightly higher contrast, brighter highlights(now overwritten with promoted version, see comment below the closing statement.)

MDD4696's long-awaited version is not ready, and since he's had some problems, he said to go ahead with the finalizing.

N.B. A few people have asked for a slightly larger (but not huge like the above renders) version of this image. I've rendered a 1280x1024 version for those people, available here. It's not really meant for voting, as I don't *think* it's got anything special about it; if you do find it visually pleasing feel free, though :P Nippoo 10:25, 20 April 2006 (UTC)[reply]
Sorry Nippoo, but that's waaaay too light on my calibrated monitor... --Janke | Talk 06:17, 21 April 2006 (UTC)[reply]
I agree... probably nothing not correctable with some balance correction. But let's not pick nits; I'm sure this rendering was for the experience of rendering it, not necessarily for improvement. After all you could easily scale down one of the larger versions for a desktop. BigBlueFish 15:32, 21 April 2006 (UTC)[reply]

Since this image is already chosen for FP, please only state your preferred version below, and if possible, the reason for your choice.

Maybe we ought to remove the dice, too? ;-) Technically, #4 really shouldn't be here - it was not the version that received 100% support - the ice cube was changed after the nom was suspended... But, if the new ice cube does get a lot of support, I could transfer it to version #5, if no-one opposes that. --Janke | Talk 08:00, 19 April 2006 (UTC)[reply]
  • Prefer #1; the version that got the support. I don't see anything wrong with it. #5 is my second choice (unless the ice cube gets imported). Ice cube #4 looks like a cube of water rather than a cube of ice... talking about fake. Besides, the changed versions weren't the ones that got the initial wave of support. As soon as changes were made, support dropped. I think that initial support should be counted. We don't need this extra poll. - Mgm|(talk) 08:07, 19 April 2006 (UTC)[reply]
  • Prefer #1 - Mgm is right that it's the one that got the support in the first place. I also prefer the light balance of this one to the others; the slightly high contrast I think adds to the beauty of the image; #5 doesn't have enough dark areas and looks just a bit more artificial. As for the ice cube, after lots of flipping back and forth between the two versions I decided the original works better. Neither look more like real ice than the other, more like different sorts of ice (leave a large smooth ice cube to melt for a little and you get the original, some machines give you ice cubes with the rough surface of the new version). The original looks more in place in the scene; the new even stands out a bit too much due to the roughness. I'd rather see the new ice cube than the new colour balance though. BigBlueFish 20:34, 19 April 2006 (UTC)[reply]
  • Prefer #1 The light is noticeably better, especially on the tall thin glass to the right (looks like a champagne glass but it appears to have wine in it). Staxringold 14:42, 20 April 2006 (UTC)[reply]
  • #1. The clipping just "works" somehow. --Dante Alighieri | Talk 18:17, 20 April 2006 (UTC)[reply]
  • 1, seems the most realistic, at least on my monitor. |→ Spaully°τ 23:24, 20 April 2006 (GMT)
  • 4, which is identical to 1, but for the better ice-cube. ed g2stalk 12:08, 21 April 2006 (UTC)[reply]
  • Support #1. The ice cube on #4 looks like it has had more of a chance to melt, but there isn't any water collecting at the bottom. The ice cube in #1 looks fresh, and there isn't any water underneath it, which is logical. I also prefer the stronger highlights.. though they appear blown, that's how the human eye interprets light from the sun. drumguy8800 - speak? 21:11, 21 April 2006 (UTC)[reply]
  • 1 The modified ice cube looks dodgy. --Fir0002 www 08:14, 23 April 2006 (UTC)[reply]
  • Prefer #1 Mikeo 10:08, 23 April 2006 (UTC)[reply]
  • 1 A lot could be done to improve it of course, but in any case I prefer the higher contrast of this version. --Gilles Tran 09:05, 24 April 2006 (UTC)[reply]
  • Prefer 5 or a synthesis of 4 & 5. The highlights seems to glare in #1. Let's not get hung up on procedure... the point is to select the best possible image among a set where any one would have passed FPC voting. Future viewers are not going to care whether they look at the "original", but they may care about the ice cube and the highlights.--ragesoss 16:34, 24 April 2006 (UTC)[reply]
  • 5. Better highlights. It's not unrealistic for an ice cube to look like that, especially if it's wet. bcasterline t 01:05, 25 April 2006 (UTC)[reply]
  • Support 5 number 1 has terribly blown out highlights (the rest have at least some blown out...), I can't belive anyone is supporting it. -Ravedave 14:36, 25 April 2006 (UTC)[reply]


Promoted Image:Glasses 800.png ~ VeledanTalk 17:26, 25 April 2006 (UTC)[reply]

(Just a note: The original file was temporarily ovewritten with # 5 during the poll, but is now overwritten with # 1, the one promoted. --Janke | Talk 05:45, 26 April 2006 (UTC))[reply]