PDA

View Full Version : my FR first encounters


Opelfruits
11-25-2005, 10:04 PM
its super quick for single objects with no environments but im having a little trouble with anything indoors. The shader tree is AWSOME, its like being back in XSI.

http://img107.imageshack.us/img107/530/chairs5du.th.jpg (http://img107.imageshack.us/my.php?image=chairs5du.jpg)

dont be fooled by this second thumbnail, the thumbnail looks pretty gd
http://img35.imageshack.us/img35/6578/render13np.th.jpg (http://img35.imageshack.us/my.php?image=render13np.jpg)

i tried using sky portals but didnt have much luck, if anybody wants the second scene to have a play with i will put it online.

K.

Per-Anders
11-25-2005, 10:23 PM
with interior scenes it's generally best to not add portals etc to the ambient tree (done via the fR light tag) as this will frequently result in splotchy gi especially the the light has a low number of samples in the light. to smooth out gi you cn increase the sample blending or increase the sample accuracy (and possible if htat fails in the advanced tab increase the actual samples). sometimes the other gi solutions are better than AQMC, for instance i sometimes find fR Image to be a better solution for stills at least.

Ernest Burden
11-26-2005, 01:13 AM
i sometimes find fR Image to be a better solution for stills at least.

Have you done any side-by-side tests of them yet? I've not had the time, but having the two other GI modes suggests that they may be good for something.

Continuumx
11-26-2005, 01:42 AM
its super quick for single objects with no environments but im having a little trouble with anything indoors. The shader tree is AWSOME, its like being back in XSI.


dont be fooled by this second thumbnail, the thumbnail looks pretty gd


i tried using sky portals but didnt have much luck, if anybody wants the second scene to have a play with i will put it online.

K.

Very good samples, I want to see more!

xeno3d
11-26-2005, 02:48 AM
Nice images. Could you post that second scene?

Opelfruits
11-26-2005, 07:20 AM
when i increased the samples it became even more splotchy and the render time shot up. i tried with the FR Image mode which was faster but still splotchy.

i found something quite funny is the read me, its says "we tried our best to make the installation as easy as possible" - hmm

http://www.krisart.co.uk/testscene.c4d

lllab
11-26-2005, 12:23 PM
i also use fr still for my current stills.

i like the edge detection prepass, and it is even a little faster with keeping high Gi quality. for some scenes it even looked better than adaptiveQMC.

cheers
stefan

Opelfruits
11-26-2005, 03:27 PM
Been testing all day, sorry these tests are not very structured, i will do some better tests soon and analyse them.

the thing i find annoying about the physical sky is the sunlight white, so i used an infinite light instead so i could change the colour.

the scene has physical sky, an infinite light with area shadows, slightly yellow. rendered on dual xeon 2.8, 2 Gb RAM. test images have no AA and are rendered at 400 x 300. also behind the camera there is a fairly big window to allow some extra light in.

bigger renders have AA - all default aa setting apart from changed to catmul rom. they also have a FR post bloom effect.

im pretty happy with the renders, now i just need to poplulate the scene.

Render 1 ::
http://img418.imageshack.us/img418/3667/32wx.th.jpg (http://img418.imageshack.us/my.php?image=32wx.jpg)

Render 2 ::
http://img418.imageshack.us/img418/4628/46qi1.th.jpg (http://img418.imageshack.us/my.php?image=46qi1.jpg)

Settings for final renders ::
http://img487.imageshack.us/img487/2788/bedscenesettings6rg.th.jpg (http://img487.imageshack.us/my.php?image=bedscenesettings6rg.jpg)

Tests ::
http://img371.imageshack.us/img371/6792/resultspage10tm.jpg (http://imageshack.us)

http://img371.imageshack.us/img371/5157/resultspage28jg.jpg (http://imageshack.us)

http://img371.imageshack.us/img371/7533/resultspage31ei.jpg (http://imageshack.us)

http://img371.imageshack.us/img371/6352/resultspage46ul.jpg (http://imageshack.us)

http://img371.imageshack.us/img371/4328/resultspage51bl.jpg (http://imageshack.us)

http://img371.imageshack.us/img371/3493/resultspage68qs.jpg (http://imageshack.us)

AdamT
11-26-2005, 03:41 PM
Nice tests, thanks!

I had a go at your previous scene:

http://www.3danvil.com/fR-2/9m24s.jpg

This was a distributed render, so figure it would have taken around 16 minuts on my X2 machine only.

I made quite a few changes, including:
*added sky portals;
*used lightmaps instead of brute force;
*used Catmull-ROM AA;
*used GI falloff;
*Color mapping;
*GI contrast;
*various other bits and bobs.

Scene file: http://www.3danvil.com/fR-2/testscene%5B1%5D_AdamT.c4d

nycL45
11-26-2005, 05:13 PM
Great intro/educational piece for us Mac users. Thanks for that great effort, Opelfruits.

Opelfruits
11-26-2005, 05:41 PM
thanks for rendering the scene adam, i did a very quick test with your settings to find out what difference sky portals make.

the top image has sky portals and the bottom one doesnt, the render times were 8 minutes with sky portals and 1 minute without. even though the portals do make alot fo difference i dont really think they are worth an 8 times increase in the render time.
http://img116.imageshack.us/img116/2621/adamt3ex.jpg (http://imageshack.us)

AdamT
11-26-2005, 05:45 PM
In this case I'd say they're worth it, as the corners look pretty bad without'em. Maybe replacing the individual portals on the diagonal windows with just two portals--one on the roof part and one on the vertical wall--would speed things up some without degrading quality too much.

Opelfruits
11-26-2005, 07:40 PM
adam, your scene rendered on my computer took 25 minutes, i got it down to 8 minutes.

made a big sky portal for the top and one for the sides like you suggested, i changed quite alot of the settings, the main time savers were the big sky portals, and the advanced settings that i changed.

adams scene rendered on my computer ::
http://img258.imageshack.us/img258/8475/adam1hg.th.jpg (http://img258.imageshack.us/my.php?image=adam1hg.jpg)

my new version ::
http://img258.imageshack.us/img258/8259/mine4lf.jpg (http://imageshack.us)

here is my new version www.krisart.co.uk/testsceneNEW.c4d (http://www.krisart.co.uk/testsceneNEW.c4d)

tcastudios
11-26-2005, 09:50 PM
Seeing these cromeballs acting as mirrors I have a humble question how reflections are calculated in fr.
I assume there are no lights, only GI in the scene.

Looking at the closest sphere, the wall "behind" the camera is somewhat lit by the GI.
Looking at the second sphere the same wall is allmost black.
Correct me if I'm wrong, but that can't be right (If the material is equal on the spheres).
A mirror only "sees" what is allready been done. In this case it looks as if the GI is calculated a second time, taken the distance from the backwall in acount.
I'd say, theoretically, no matter how far away the mirror is, the reflection should stay the same.

Since I have no fr to check this myself I'll check in Cinema later.

Cheers
Lennart

Opelfruits
11-26-2005, 10:01 PM
the reflections seem fine to me, i think the reason the back wall seems darker in the further away balls is because this part of the reflection is much smaller.

if you rendered the image much bigger to see the reflections in these balls i imagine the illumination of the back wall would be more clear

AdamT
11-26-2005, 10:07 PM
Good job getting the time down. Your version looks very good!

tcastudios
11-26-2005, 10:23 PM
Opelfruit, please don't missunderstand me, it's no bashing. I'm only interested in how it works. Given that fr have alot of parameters it might be settings for this as well.

But ,taken into extreme (given how it looks to me) if only the backwall was lit up and a cromesphere is put far away in the "corridor", no reflections would be seen in it.

I would imagine a "true" reflection would be a "baking" of the environment seen from the mirrors (sphere's) point of view that is then mapped onto it.

And , again, just wondering, but it struck me as the first thing looking "odd" in the picture.

Cheers
Lennart

Per-Anders
11-26-2005, 10:39 PM
no baking is involved. it's just normal raytraced relfections as you'd expect. simply put your eyes are being decieved.

it's because of the contrast that it looks this way. the balls further down are relfecting a lot more of the more brightly lit room, so the smaller darker patch looks darker than it is (it's much smaller and in much closer proximity to the lighter areas thus it's contrast is increased dramatically). if you take it into photoshop you'll see that the dark distant wall in each sphere is roughly the same color and chromanance (in fact in the distance spheres it's actually lighter than in the nearest sphere!).

tcastudios
11-26-2005, 10:46 PM
Thanks. So it works as it should!

Cheers
Lennart

CGTalk Moderation
11-26-2005, 10:46 PM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.