PDA

View Full Version : Rendering an oversized frame


Simon Wicker
09-08-2008, 10:02 AM
One of the neat things available in renderman is the easy ability to render out a 'padded' or oversized frame - simply put the ability to render extra area at the top and sides of your frame so you have more image to work with.

Matte painters often need this so they can add a painted extension beyond the current extent of the frame.

Also if you need to add lens distortion to match the cg render to live action you will often need to warp in information that may be outside the initial frame size.

Other times you may want to add in camera shake in post, once again you would need to have 'extra' data outside the frame so you don't just see black creeping in.

In cinema it is easy enough to add canvas top and bottom (in the render settings you can simply increase the height values and the render compensates correctly and you render data that wasn;t visible previously.

However you can't arbitrarily adjust the Width of the render size because the width is tied to the fov of your render camera (and effectively it controls the resolution of the render - so if your intial render is 1024 pixels and you change it to 2048 your fov stays the same, there is no change to what you see through the camera but you have doubled the resolution).

What you really need to do is to create a new render camera, adjust the render settings to add canvas and then adjust the fov of this new camera to fit your old render size into the centre of the frame.

I normally just pull out a calculator to work this out but then I also realised that I could also use Xpresso to work this all out within cinema. I've enclosed a scene with my little padded render calculator gadget. Just click on the null and look at the user data to see what is what.

So, the problem and how it is solved. Lets just say you have a render and you need to have 200 pixels of extra canvas all around because the compositors want to have a bit of camera shake added in post.

Your current render is 1024 x 576 (i.e. PAL widescreen) and your current render camera uses a 28mm lens.

Your target render settings are 1424 x 976. The 976 bit is just fine. But you need to know how much wider your camera needs to be to fit the old plate into the new one. So, click on the padded render camera user data and start filling in the information. Once you have filled in the first three values the Xpresso calculates that our new camera focal length is 20.135mm.

So if you duplicate your original camera, adjust the focal length to 20.135mm, set the render settings to 1424 x 976 you will render a sequence that is the same view as your original camera except that it has an extra 200 pixels added to the canvas.

Cheers, Simon W.

JustinB
09-08-2008, 10:08 AM
Thanks Simon

A Very useful expression. Much appreciated:-)
Cheers
Justin

chris_b
09-08-2008, 01:52 PM
Thanks Simon this is very handy! Another frequent usage is for people adding Motion Blur in post using Reelsmart. When objects enter or exit the frame (especially fast moving ones) Reelsmart tends to smear/tear the edges. By adding an overscan these artifacts can be kept out of the frame. :)

JoelOtron
09-08-2008, 03:48 PM
Thanks so much Simon.

metagrey
09-08-2008, 06:03 PM
Thanks for the clear explanation and the file.

colibert
09-16-2008, 01:59 PM
thanks for that, very useful indeed :cool:

CGTalk Moderation
09-16-2008, 01:59 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.