vray for xsi

[joconnell]

Again no idea - still haven't gotten around to a job in XSI yet myself, I'm trying to stay away from it at the moment, I've enough stuff booked in

[Morbid Angel]

studioDIM, rapid seems like a good motion blur, but its a false cover over a really bad method underneath.
There are a number of things rapid doesnt support:
Lets start from frame buffers. If you use mental ray's frame buffer output to generate passes, rapid doesnt antialiase them at all.
Rapid doesnt work with transparent objects very well, doesnt work with reflected objects verywell. For example if you have a chromatic sphere which is reflecting an object, and its spinning around its axis, in theory the reflection should not motion blur because its constant, but rapid will motion blur it. This is because rapid samples everything only at one time, before motion blur begins, then it blurs it, where are mr scanline samples it at every frame increment. Vray does same however its much better with adaptation of samples and smarter, then mr just brute force samples every increment as it was a whole frame, which is why it takes so long for default mb to render.

[studioDIM]

Nice explanation, Dmitry.
Couldn't have asked for any better myself, lol

And Jo, booking is fine, it's doing it all which is the hard part...

Lele

[Morbid Angel]

Well I couldnt resist and did some tests, the results are stunning:

Below is a set of objects, there is a chromatic sphere which is rotating around Y axis, infront there is a semireflective box spinning. Two area lights, and blue environment, no gi. The motion blur geometry samples is set to 6.
As you can see vray did a correct trancing in every count; the rotating sphere is not motion blurred, the motionblurred spinning box is reflected correctly in the sphere, and is also casting a correct motion blurred shadow. Render time was 27 sec.
Next is mental ray scanline, the sphere has been motion blurred you can see that judging based on the reflected objects in it. You can also see that the spinning box has been blurred but it also seems its not been sampled 6 times. The render time was 1:55 min
Next is mr rapid, you can see that its also blurring the reflected sphere, its not tracing the shadow on the ground correctly from spinning box, and the background color is not traced correctly at all, also the spinning box blurred edges are not showing through correctly against the objects behind. The render time was 0:56 min.

[GeneCrucean]

Nice test bud.

THIS is why I want *read thread title*.

[studioDIM]

aaaah, yes, nice one bud.
I have it a lot more clear now, thankyou
I thought it was mighty fine, but i see how i didn't cover all areas of it in detail, concentrating on deform-blur rather than shading features (which i stupidly assumed were correct...).
Well, comeon vray then

Lele

[yog]

With version 6, the SDK access to our entire suite of rendering interfaces including the render tree, shader balls and interactive render region can all be easily accessed. Studios interested in using rendering packages such as Renderman or Brazil will have the capability of integrating these tools into XSI.

http://www.softimage.com/products/xsi/default.aspx
http://features.cgsociety.org/story_...referer=cgnews
http://www.xsibase.com/news.php?detail=1893

I admit, when I saw this in the XSI-6 press release, V-Ray was the first thing I thought about.

[vlado]

I hope it's not the same level of "access" that was added to Maya...

Best regards,
Vlado

[jacksc02]

What do you mean vlado?! Surely Autodesk has made things easier?

[vlado]

Next is mental ray scanline, the sphere has been motion blurred you can see that judging based on the reflected objects in it. You can also see that the spinning box has been blurred but it also seems its not been sampled 6 times. The render time was 1:55 min
Next is mr rapid, you can see that its also blurring the reflected sphere, its not tracing the shadow on the ground correctly from spinning box, and the background color is not traced correctly at all, also the spinning box blurred edges are not showing through correctly against the objects behind. The render time was 0:56 min.

Does that mean that you can't actually get correct motion blur for the sphere at all with mr (I tried mr with scanline on/off and rapid scanline and they all blur the reflections in the sphere too)?

Best regards,
Vlado

[Morbid Angel]

vlado, indeed you cannot. Infact the most common method of motion blur in mental ray in 99% of productions, which include movies like poseidon, used motion vectors for mb. But with motion vectors you cant get this result eather. So as it stands right now, I think mr's blur is in pretty bad shape. And we have to suffer because of it on every shot.

[Morbid Angel]

Here is a reply I got on mental ray forum:

This question came up on the mailing list forum, and we addressed it in training class recently.

In 3D motion blur, the shading happens after objects are moved and therefore, the renderable polys will be moved before the material shader traces reflection rays. However, one optimization is that the normal is not transformed.

So a mirror that moves in translation will render without blurred reflections, but one that rotates could exhibit blurred reflections.

We wrote a shader that could adjust the normals appropriately for a spinning sphere. You'd run this shader before any illumination shader for the material, much like a bump shader.

[GeneCrucean]

Vlado, what needs to happen for this to become a reality?

[GeneCrucean]

GREAT!! Nothing needs to happen! Word... I'll be expecting an announcement soon then.

[Morbid Angel]

hahahah....patience - will be rewarded.