Overscan for VFX

Overscanning is a process where extra pixels are rendered around the intended resolution (not like Eevee's overscan). That can be useful not only for lens distortion workflows, but also post camera shakes, better glows at the edges etc. With a bit of math it's pretty simple, but why not have it as a feature and be future proof? Even .exr files support some form of overscans, maybe that could be implemented as well in the future.

Here's a mock-up of a possible UI:

In some cases, x-axis property would be enough, Y could be calculated in proportion to the aspect ratio, since lens distortions generally transform the pixels radially from center, but some softwares use the same overscan for X and Y, also by using proportions or a simple % slider, decimal values would be introduced, which is not good - pixel perfect accuracy is needed. That's why 2 integers would be a better option.

This feature should be independent of the render engine, all it needs to do is increase the render resolution and FOV so the cropped image exactly matches the original. The new resolution would be:

resolution_x + overscan_x * 2
resolution_y + overscan_y * 2

And the easiest way to increase FOV without changing any other lens characteristic is to internally multiply Sensor Size by one of the factors below this image:

(resolution_x + overscan_x * 2) / resolution_x    // new x res. div. by old res.
(resolution_y + overscan_y * 2) / resolution_y    // same but for y

Now which one to use depends on Sensor Fit in camera settings. X values for Horizontal , Y for Vertical . The Auto sensor fit is a bit more complicated, since it should probably take the altered resolution into account. The problem rises when the aspect ratio is close to 1:1 where the overscan can change the overall orientation. But basically if the image turns out to be horizontal, then again use the X values and vice versa.

There are 2 main ways to visualise this in the viewport (camera's point of view). Either show a dashed line outside the set FOV and change the settings only for render, or already show the wider camera with the rectangle inside. Both have their advantages but I'd opt for the 2nd. Here's a comparison:

There's probably more things to consider - Render Region , Background Images (which should have an option to align to both sizes), maybe if anamorphics would work the same, how to remind people of the final/different resolution etc. but I feel like this is long enough...

While I was making the UI mock-up I also changed the Dimentions tab a bit (e.g. Pixel Aspect , Time Remapping ), it's more or less just my preference, but take a look at it, maybe some of it makes sense to more people then just me, like why do we have 2 values defining Pixel Aspect Ratio ? Also one small UI nitpick - the Frame Rate setting doesn't align properly, there's a bigger gap between the text and dropdown, which seems to be 4 px narrower. Other instances of the dropdown menu don't have this problem.

Anyway, I hope my English is OK, this is my fist proposal here, so thank you for reading!