ypoissant

You can blur the HDR in HDRShop. That might help and this should be more realistic anyway with the reflections on the car.

It would help if you supplied a picture of that. But more generally, HDRI is a difficult beast to master. Getting pixelation is the nature of the beast. Why? Suppose your open sky if 1000 brighter (it may very well be even brighter than that) than the rest of you HDRI scene. If even 1 out of 256 samples falls on the sky, then this illumination value will very probably saturate the pixel even after averaging. Oversaturated pixels produce aliasing.

(I assume you are refering to the exposure of the HDRI environment map) No. You have to modify the exposure in HDRShop. Load your image in HDRShop, change the exposure with the numeric keypad + or - keys. Once you're satisfied with the resultant exposure, press Ctrl-0 (that's zero) and save.

Pick the Uffizi HDR light probe from Debevec Web site. Load it in HDRShop and transform it into a Lattitude-Longitude map and save it back as an HDR file. Then use HDR2EXR to convert this HDR file into an EXR file. Now you have a spherical environment map for A:M. Method 1: Start A:M and load that Uffizi EXR file into the PWS Images folder. Create a Material, name it Uffizi and change its attribute to Environment plugin. Open the Environment plugin properties and load the Uffizi EXR as the image property. Drop the Uffizi material onto your car model. Method 2 (the one I used with those reflective spheres): Start A:M and load that Uffizi EXR file into the PWS Images folder. Create a Material, name it Uffizi and change its attribute to Projection plugin. Open the Projection plugin properties and load the Uffizi EXR as the image property. Import a sphere model and drop that Uffizi material on the sphere object. Set the sphere ambiance to 100% and its diffuse falloff to 10000% so that it is not affected by the lights in the scene. Drop the sphere in your scene and scale it large enough to cover the whole scene including the lights in the scene (at least 10 times the size of your car). Set your car reflectivity to some low value like 10%. Additional note: Orient the Uffizi sky opening to align front to rear with your car. This may look like a trivial details but this sort of details is important when lighting a car.

Not to make her more realistic. But a little color variations in her skin would look just a little more natural. I think your character does not need to look realistic. You have a nice style. Only more natural skin. Right now, because the skin color is so uniform, she looks like she is made of rough plastic or vinyl. Also did you try the Skin shader on her skin? That "nude" render reminds me of a surrealistic painting.

The Uffizi is one of the HDRI probes available from Debevec Web site. I'm pretty sure I recognize it in the car reflections. Transparency is caped but reflection is not caped. At least I don't recall it being caped.

I see that you are using the Ufizi light probe as the environment map. Did you convert it to OpenEXR or to a format such as TGA? Did you try to map an OpenEXR of this environment?

Cool (or should I write "hot") santa-claus you got there. I don't beleive in Santa Claus anymore but I would be ready to change my mind. There are two nit-pick that are catching my attention: The shoes don't have a sole. This looks funny. The skin looks too uniform in color and thus look unnatural. A color map with color variations would help. Legs and thigh are generally lightly bluer with legs bluer than thighs (this increase with age but your model is rather young). Arms are bluer too but the hands are generally lightly yellower. Also, a little bump maps to get a little bit of muscle definition would also help get a less uniform look.

OK. I see. Glad you found a solution. BTW, When I was working on designing my skylights, I also did a few of those experiments that turned not quite like what I expected. But those were good learning experiences.

Rodger, I'd be currious to see how your skylight was setup to get such a result. I've been using skylights for years without getting this and the lower lights of my skylight rigs always extend below the horizon.

The illumination still looks like it is under a cloudy sky. I would suggest reducing the sky intensity and increasing the sun intensity first to get nice sunny-day shadow contrast first. Then, see if you need to increase the yellow/blue contast again. Also, that noisy shadow banding coming from the sky is disturbing. Set your skylight to cast 2 rays. Even with your skycast setup, that should help reduce those bandings.

Right. You just qualified the rule: "for such a small room". What I mean is statement like "photon samples shouldnt be higher than 100" need to be qualified. Otherwise it just looks like a blanket statement. In this case, since the photon samples allow values up to 500, one would wonder "why 100?" Usually, yes. One situation where one would need to increase the photon samples is when a scene produces light leaks: bright spots or bright trails of luminosity when two objects are very close to one another. That happens because several photons are reverberating between the two nearby objects producing an accumulation of photons in those areas. Because there is an accumulation of photons in those areas, the "100 samples" will be reached quite rapidly and within a very small area, much smaller that the specified sampling area. So to get rid of the light leak, one need to increase the photon samples in such a way that the accumulated photons contributions are sinked into the other photons in the area. Photon Sample Area and Photon Samples work together but the first met condition will stop the sampling. So even if you specified a large sampling area, if the number of samples is met before the sampling area is fully sampled, the sampling process will stop there. Yes, it is. A good rule is use the smallest photon samples that will produce the acceptable result in the minimum of time. The photon sample calculator is a nice helpfull tool. But still just a tool. It will only give you a very good starting point. Most of the time the starting point is good enough. But in some situation, one may need to find more optimal settings based on issues that may arise while rendering.

You need to qualify this rule. This is not true in all circumstances. So in which circumstances does this rule apply?

And also, if you use the desaturation trick, lower the overall contrast in the image. All this can be done with A:M post-processing plugins. And while we are there, don't also forget to adjust the image gamma too.

I agree. This is way too dark to figure anything usefull. You must simulate darkness without having such a dark image. To simulate darkness. you must desaturate all your colors. In night vision, we almost don't see any color. Almost only gray tints. And we cannot distinguish details. A coarse film grain will do that. By using completely desaturated colors and a coarse film grain, you will be able to increase the general illumination in your scene.

This is looking great so far. You models are very well executed. Take care of your color saturations. The chair blue tissue is way too saturated. And the wall orange-brown is also too saturated which cast a general orange tint to the whole scene.

Ed, Thank you. Really, when I come on the forum and see users producing those nice renders with A:M radiosity, like you did, this is enough to make my whole day. Keep up the good work.

Nice scene and render. It feels real. As Vern mentioned, you have several places with pure saturated colors. Those immediately jums at the eye and give away the CG nature of the image. Finding a good balance might be tough. Take a look in the "Lego" thread in the Radiosity forum where I posted a few normally saturated colors RGB values. FYI: The size of the models and the scene have nothing to do with render time. What influences render time is the number of lights in a scene, and for radiosity: the number of photons and the number of photon samples and the number of final gathering samples.

That is the effect of light falloff on default plastic material. For plaster ceilings painted with matte paint, you should try the Oren-Nayar shader. You should take a visit to the Cornell Box tutorial in the Radiosity forum.

I wouldn't attempt any serious animation with radiosity unless you already have a render farm available. This will lengthen your render time considerably. And seriously, given the general dark mood of your scene, I don't even think radiosity would be really helpfull. You would be better off with getting the mood with judicious placement of additional fill lights.

Cool image. Another variation of the Sierpinsky pyramid but your variation is more interesting because it better fills the space.

Very good modeling job. I'm impressed.

Cool image. I love the lighting and the iamgination for that world design. Bravo!

Personally, I wouldn't put this in a demo reel. Think of the implied message. Something is falling appart. What is it? It is not the logo. It is whatever is written in the logo. If you really want to use physics I'd suggest something that will convey the inverse message. As an example, let's say your logo is sitting on a pile of blocs. Than a large boulder falls on the ground and shake the whole scene. All the blocs falls apart but the logo, after shaking a little stays firmly there.

First, and by far the most important aspect to consider, is your car needs something to reflect. With nothing to reflect, you could tweak forever without getting anywhere. Setup a sky hemisphere onto which you will project an environment map and set its ambiance to 100%. Then we can start looking at the other adjustments.