Sometime ago I've seen on a book a striking photograph of the Auiguille Noire du Peuterey under a strong wind, in the Alps. There was a sort of cloud formation streaming over the major crest, it was really marvelous and amazed me.
I thought about it for a while then I started to think how could I reproduce something similar on TG2. It was quite difficult because TG2 (almost actually) doesn't manage 3d clouds and cannot get a very accurate control over cloud placement. I'm not so familiar with functions, so I decided to utilize image mapping: I rendered many orthogonal projections of a black and white cloud density shader at high resolution, then remodeled them in Photoshop to give every layer the shape I wanted. The concept was to create a cloud formation with an accurate '3d volume' that follows the exact model I had in mind, and to do this I overlayed multiple cloud layers in the scene, like a cross section of an object. Something like that:
(http://img379.imageshack.us/img379/4194/layers2bp8.jpg)
The shapes were modeled directly on Photoshop over an orthogonal projection of the terrain, then I placed them with the right coordinates in the scene, as density shaders for the clouds. There are six cloud layers in total.
And this is the final result:
(http://img524.imageshack.us/img524/1910/finaleprocessedoz8.jpg)
Well, maybe doesn't seem anything particular, but I worked a lot on it :)
I'm trying to do many versions in Photoshop with different contrast, colour balance, little corrections etc. I'm undecided and can't manage to get a final version that completely satisfies me. I'm not an expert in post processing so I have to experiment a lot.
Any comment, hint and critic is very welcome, as always!
Regards,
Lucio
Wonderful! I though It was a photo at first.
Lucio, magnificent!
HOLY FREAKING COW!!!!!!! :o I don't know what to say! That is without a doubt one of the most amazing TG renders ever made.
Wow this is amazing, I thought I was going to scroll down past your reference shot to find the terragen one! Faultless! :D
I'm actually working on a scene very similar to this at the moment but yours is way ahead of where I'm at. I like your image masking technique, I had thought on similar lines but never got around to testing the theory.
Richard
Very nice. These clouds remind me of some stuff I saw while hiking in the alps. The landscape is very realistic, but I think you could really go far with the clouds here by adding more detail. This is a great render so far, though.
Did some quick post, hope you don't mind.
Oh no, thank you Will! it's an interesting version. I encourage any other post-pro effort over my work :)
o_b: I agree with you, I already noticed that while the final render was completing. Maybe I've also increased too much some edge sharpness. I'll try to do something better now.
I would like to enhance a bit the contrast between darker and brighter regions of my cumulus without recurring to post, I played with many parameters but cannot come to a real solution. Do you have any hint?
Volker? ;D ;D ;D ;D
(Don't mind, it's a question between me and him)
that's amazing !!!
it really looks like a real picture !
congratulations for this !
Come on, Lucio - if you don't stop raising the bar no-one will ever catch up.
;D
Makes me want to go climb something. :)
Nice Lucio!
I wish i understood the factors you use in things like this :'(
Again Really well done!
Nice idea and result. The only thing I would say is that perhaps the pinkish colouring in the clouds is a little strong.
The clouds and landscape are amazing, but can nobody else see the banding in the sky?
Kevin.
i do.
I do thats why I did the post and blurred it out.
Quote from: Kevin F on September 16, 2007, 01:38:59 PM
The clouds and landscape are amazing, but can nobody else see the banding in the sky?
Kevin.
Well, I can see the bands very clearly. I attributed the cause to the bad old monitor I'm using now, the other I have blew out a week ago.
Will: I didn't noticed you blurred them because your postprocessed version was quite dark
Good work!
You could use your Layer 1-imagemap and with each higher cloud-layer decrease the edge sharpness ... this could be nice.
Great work Lucio ;)
Here's an attempt to improve the work
I've re-rendered the ice parts with higher GI to get a better appearance and linked them to the original image with post, added a little glow somewhere over the clouds and tried to remove the banding in the sky. I'll try to re-render clouds with improved details and see what happens
(http://img77.imageshack.us/img77/4858/finaleprocessedcopiagv5.jpg)
Regards,
Lucio
amazing man.....no words....
I hope you don't mind, I brought it into photoshop and had fun for about 30-45 minutes!
(http://img219.imageshack.us/img219/2728/finaleprocessedoz8bpit8.jpg) (http://img219.imageshack.us/my.php?image=finaleprocessedoz8bpit8.jpg)
If you would like the .psd I am more than happy to give it to you, since you are indeed the original creator of the work :D.
EDIT: I had a color filter switched off, so the colors were a bit off... Updated image.
nice alternative......
BPauba, this is very nice at the top, but I like the ice at the bottom on the original.
Quote from: calico on September 18, 2007, 04:28:52 PM
BPauba, this is very nice at the top, but I like the ice at the bottom on the original.
The ice was nice, but compositionally I do not think it works. It really catches the eye and struggles with the top of the mountain for the focal point, so that was the reasoning behind taking it out. :D
Hi PBauba, I can't see if you have attached an image or what.. I only see your post! Can you send it in a private message or via mail? Thanks :)
Hi Lucio.
I have been playing with cloud masking on and off and was going to start on a similar project with the alps. Even the landsat data for the area still has orograhic clouds along one side of aridge through a significant portion of the image. The main problem I find is that as soon as you start masking the cloud fractal you run into problems with the edges smoothing out too much.
My next set of experiments in this area were twofold: 1) play around with order of nodes using blending shaders. e.g. fractal blends image or image blends fractal.. there are differences in the outcome but I haven't done enough testing to fully grasp it yet... 2) breaking up up the non-white/black portions of the image map output (i.e. after anti-aliasing and colour adjustments) with a fractal noise. Sounds simple enough, although the interactions betwen the two sets of fractals (breakup and cloud) complicate things a bit.
Hope this makes sense... ;)
You're definitely heading in the right direction though :)
Thanks for explaining.
Quote from: BPauba on September 18, 2007, 05:44:28 PM
Quote from: calico on September 18, 2007, 04:28:52 PM
BPauba, this is very nice at the top, but I like the ice at the bottom on the original.
The ice was nice, but compositionally I do not think it works. It really catches the eye and struggles with the top of the mountain for the focal point, so that was the reasoning behind taking it out. :D
Quote from: bigben on September 18, 2007, 07:02:09 PM
Hi Lucio.
I have been playing with cloud masking on and off and was going to start on a similar project with the alps. Even the landsat data for the area still has orograhic clouds along one side of aridge through a significant portion of the image. The main problem I find is that as soon as you start masking the cloud fractal you run into problems with the edges smoothing out too much.
My next set of experiments in this area were twofold: 1) play around with order of nodes using blending shaders. e.g. fractal blends image or image blends fractal.. there are differences in the outcome but I haven't done enough testing to fully grasp it yet... 2) breaking up up the non-white/black portions of the image map output (i.e. after anti-aliasing and colour adjustments) with a fractal noise. Sounds simple enough, although the interactions betwen the two sets of fractals (breakup and cloud) complicate things a bit.
Hope this makes sense... ;)
You're definitely heading in the right direction though :)
This is very interesting bigben, would you like to consider some cooperation? ;D
I must say that I've focused my mask manipulation experiments mainly on Photoshop. Adding fractals in TG2 could be the next step. In fact cloud masking is quite labourious but can give a number of very good possibilities. I'm looking forward for your next works :)
Two heads are better than one ;)
I'll have a go tonight at a clip file for breaking up the border of an image map and post that for you to test. I suspect the hardest part will be tweaking the fractal used for the breakup.
I'm guessing that's the Matterhorn. I may have a TER file lying around somewhere I can use ...
Great, let's try!
OK, I got very distracted last night just trying to get close to where you are.... Let's just say that I now have a far greater appreciation of what you have achieved so far ;D
That left my morning train ride to refocus, and unfortunately I forgot to save often ;)
Nevertheless I did manage to split up the image map into it's white and non-white components so that we can start to tinker with the border of the image map without affecting the rest of it. I'm not entirely convinced that this is going to be a big fix, and will probably be only useful in situations like yours where your clouds are fairly close to being what you want.
The biggest problem with masking clouds is the thickness of the cloud at the edge of the mask. If it's almost full thickness then you have a very odd looking cloud. The only way around this might be to break up the border with another fractal of smaller scale but this will still be of limited use... If however you're relatively close to having the right result then the additional border breakup may provide a subtle improvement that makes just a bit more believable...
I'll attach this morning's work although all it does is isolate the different parts of the image map. The zip file contains my test image so you probably don't need it. I'll persist with this test regardless of how well it progresses with clouds because it will be useful to me for my river masks
[edit] It would be really nice if there was an additional input in the cloud node that scaled the cloud thickness based on a value from another node. That would open up a lot of possibilities for creating different cloud formations without trying to mask the cloud fractal. Easy to say... possibly not that easy to program ;) [/edit]
Thank you bigben, I'll have a look on it
This could be also a good occasion for me to learn something about functions :) I will try to apply your masking technique to some graphic result. Thank you very much!
PBauba: I've seen your image. I like very much how you worked over the bottom part, I think it's compositionally much better than mine. The superior part is also very nice but is a bit too 'illustration-like' and it's not the mood I wanted. However thank you very much and great work :)
I had a few more problems which I need to test more to isolate the exact causes. I thought I was using some fairly stanard blending of shaders but I ended up with the flat masked result at the top and bottom of the cloud layer and nothing in the middle (when viewed from below). This disappeared completely when I moved the camera above the cloud layer.... kinda weird but it may have been my fault. I'm going back to my earlier cloud masking tests (which still work) and will try and combine the two. If you could email me a clip of a single cloud layer with any masks (images not required) I can try and fit it in with what you're doing.
Quote from: bigben on September 20, 2007, 07:09:07 PM
I had a few more problems which I need to test more to isolate the exact causes. I thought I was using some fairly stanard blending of shaders but I ended up with the flat masked result at the top and bottom of the cloud layer and nothing in the middle (when viewed from below). This disappeared completely when I moved the camera above the cloud layer.... kinda weird but it may have been my fault. I'm going back to my earlier cloud masking tests (which still work) and will try and combine the two. If you could email me a clip of a single cloud layer with any masks (images not required) I can try and fit it in with what you're doing.
bigben, your technique works very good. I'm testing it in different ways and also given me some idea on how to expand it. I'm trying to go on something concrete, and soon I'll let you know :)
Hey lucio, what is your email? I can send the image right now if you would like...
Oh, thank you! You can send it at usciouscio@yahoo.com
Thanks again :)
Quote from: Lucio on September 21, 2007, 07:21:37 AM
bigben, your technique works very good. I'm testing it in different ways and also given me some idea on how to expand it. I'm trying to go on something concrete, and soon I'll let you know :)
I wasn't entirely sure how useful it would be for you as it would largely depend on the type of image you were using and the method used to apply extra breakup of the edges. Glad to hear you found it useful. :)
Your technique gave me new ideas and I've got some result. Here's a simple shader structure I elaborated for masking clouds directly in TG2. I piped in a 'subtract colour' node a density fractal and an inverted greyscale mask. Here's a test render
(http://img181.imageshack.us/img181/7842/testrenderfg0.jpg)
As you say the hardest part of the work is to get decent detail from border blending. Now, this render looks horrible because it's just a test done with a unappropriate image mask but potentially this technique could bring good results, if you got a fine greyscale tuning of the mask borders. I've alleged a .zip file with the .tgd and the image mask so you can continue the work on it if you want.
Now as a next step I would like to integrate your technique to get some extra-detail from border blending. I'm also thinking about a method that could connect surface layering with cloud masking, even if it's not very simple. Thank you! :)
Regards,
Lucio
I'll have a play with this. I was thinking along the lines of using an inverted border mask as well so that you gradually subtract more detail from the masked cloud border as you move out to the edge of the cloud
If you want to save some memory for TG to work with convert your mask image to greyscale and save it as LZW compressed TIF. The compression only saves space storing the file, but using greyscal instead of RGB will make a significant difference to memory usage which you may need for objects etc...
Also... if you're using a broad blur on your mask you might also experiment with smaller images as the internal AA applied to images in TG will act as a good blur for low res images. It largely depends on what level of fine detail you need in your mask. You can also use a colour adjust shader to counter some of the AA on low res masks if it becomes too blurred.
Thanks for your hints bigben. I'm discovering step by step the world of functions and I'm amazed by the further possibilities offered by this extremely powerful instrument :)
If you looked at my .tgd maybe you noticed that my masking technique scatters many undesired 'cloud fragments' all over the scene. I reworked on it utilizing some of your concepts and now I've got a method that completely cleans the contour of the masks. Take a look at the new .tgd I've alleged here. Here is another test render
(http://img524.imageshack.us/img524/7883/testrenderfinalaa6.jpg)
Now this masking method is complete and should work fine. It's a possibility. You can build your cloud layers directly in TG2 as you want, then mask and reshape them in all the ways you prefer. I'll try to utilize it in some of my next works. A bad aspect is that on the other side it seems to slow down significantly rendertimes. I'll do some specific test to see the effective increase.
If you find anything that could improve it or something alternative please let me know! :)
Regards,
Lucio
Hmm got a bit off topic.. maybe I should post that in the shared files section
Who cares? This is cool.
I played around with this for a while to figure out the best way to build what I was thinking of. Some approaches sounded OK in theory but didn't work... but in the end I got there (in theory) without complicating things too much. Render times are noticeably longer, but not as excessive as I have seen in previous masking experiments.
I rearranged the placement of the image map a little but not for any great reason other than that's one way I normally do it ;). The colour adjust is also not essential. I use it for tweaking the mask for testing and sometimes leave it in. The surface previews were also left in from testing. I find it quicker to look at those to get a feel for what's going on instead of waiting for the cloud.
All of that aside, the only thing it really does is subtract a second cloud fractal from the cloud, gradated from the the edge of the mask inwards. The only extra tweak then is the settings of the second cloud fractal. I hope it works.. or at least gives you some more ideas :)