Jungle - RTO version @ page 5

Started by Tangled-Universe, April 02, 2009, 11:58:14 AM

Previous topic - Next topic

Tangled-Universe

Quote from: Mohawk20 on April 13, 2009, 11:06:57 AM
I can't check it out now, rendering on both pc's. But the fact that you can share objects in a clipfile means you used the internal grass object. Nice to see those looks pretty good as well.

Ghehe...that's a mis-assumption. The clipfile still contains the path to the model, but not the model itself.
As you can see it is the first Gras model from Klas.

Martin

FrankB

This has become the great render it promised to be in the earlier partial renders. good to see this as a whole now. :-)

Do you still have the .exr? I think it can be improved a little more in post, to make the lighting a little more dynamic. I would also try a slight blur. Also you could try that LAB mode curves adjustment on a and b.

Regards,
Frank

Tangled-Universe

Quote from: FrankB on April 13, 2009, 11:51:21 AM
This has become the great render it promised to be in the earlier partial renders. good to see this as a whole now. :-)

Do you still have the .exr? I think it can be improved a little more in post, to make the lighting a little more dynamic. I would also try a slight blur. Also you could try that LAB mode curves adjustment on a and b.

Regards,
Frank

Thanks Frank :)
Of course I still have the .exr
You're really better at this than me. I did some curves in LAB- and RGB-mode. But more subtle. You'll see clearly a difference without it, but perhaps I'm a bit to reserved with it.
If you want I can pass you on the exr when you're online.
I deliberately did not blurr it because then the tiny details in the grasses will be lost. A gaussian blur of 0.2 or 0.3 radius will look fine.

Martin

Saurav

This is really great. :o

As Frank said the finished version truly lived up to the potential of the drafts.

Dune

Fantastic image, Martin. You really get sucked into the depths. Do you find much difference in detail when using 1 instead of 0.5 or so? I did some sample renders with different settings, and there's wasn't a huge amount of difference, so now my standard is 0.5 and AA 5. No GI as well, but 1 or 2 fill lights. These were also the settings for my medieval Zutphen image.

---Dune

FrankB

I'm not Martin, but let me try to answer anyway :)

Yes, there is a lot of difference between 0.5 and 1. However it depends on the level of detail that your render produces. Depending on the distance of the camera to - let say - a rock surface, you may not be able to distinguish certain rock details anyway.
Usually for stills, it's well worth to increase the render quality to above 0.5 in the final render. My personal preference would be to use 0.8 as the minimum for final renders. I'm actually using 0.8 fairly often. Up to 0.9, and if the scene details requires it, I also use 1 every now and then.

For Martin's jungle render, I reckon, 0.9 would have been sufficient IF he had used the soft AA filter. Could have saved him a few hours of render time. Now that we have a choice of AA filters, they really have a big impact on the visible detail in the render. Martin prefers the Mitchell-Netravali filter, probably because it somewhat brings out detail even stronger. As a matter of personal taste, I prefer the soft filter for renders with a lot of distant vegetation - reminds me more of how photographs look like. That filter however "destroys" some detail in the render, hence decreasing the render detail from 1 to 0.9 in the jungle render would have been acceptable.

However, once you do have a lot of potential detail, 0.5 is definitely wasting lots of it. At the very least, use 0.7.

Cheers,
Frank

Tangled-Universe

Thanks for jumping in Frank :)

A bit scary the way you seem to know me in the meantime ;)
I think you have explained it very well about choices for renderquality.

I prefer the Mitchell-Netravalli filter because compared to sharper filters and the softest filter like cubic b-spline this filter offers a bit of both. It's a nice blend :)
I agree for distant populations the b-spline would look more convincing. Then I would indeed have to lower the AA from 12 to about 9 and then together with a detail reduction to 0.9 it could indeed save some rendertime. My guess would be around 20% for certain.
However, I'm not sure about the details of the rocks then.
Therefore the MN-filter, quality 1 and AA12 seemed best for me, though not rendertime-wise :)

Martin

Phylloxera

Great work, beautiful combination of different plants. distribution was successful. Very realistic! I love this picture!

Tangled-Universe

#68
Thought it would be nice to render this one using the raytracer for objects.
Rendered at 1600x900 in 8 hours. My estimation is almost twice as fast than the micropolygon renderer would do (just under 20 hours for 1920x1080).
The lighting on the models is much better and as you can see the raytracer is much preciser as well, hence the grasses are "gone".
They are there of course, but so thin that you can't distinguish them anymore. To get them visible I'd have to increase the density twice or something.

Martin

[attachimg=#]

Henry Blewer

http://flickr.com/photos/njeneb/
Forget Tuesday; It's just Monday spelled with a T


FrankB

couldn't resist, Martin ;)


Tangled-Universe

Quote from: FrankB on January 04, 2010, 08:29:02 AM
couldn't resist, Martin ;)



I don't blame you ;)
A tad bit overdone, but nice indeed :)

Kadri

Nice  :)
It is a dramatic render time improvement as well like a new powerful cpu .
Planetside should emphasize this.

Kadri.

FrankB

I also think you should've rendered this with higher AA settings.