How to animate the render size?

Question

I have a lot of camera background images for reference in an architectural project. I made an animation where, for each frame:

• The camera position is defined, corresponding to one of the images
• the same image is used as viewport background. Images have been grouped in a video, and the video is used as a camera background instead.

This way I can change my reference image and match the point of view by just changing the current frame.

However some of the images are in portrait mode, some in landscape mode, and the render size ("resolution") has to be switched accordingly, frame by frame.

This works manually, but I'd like to do this automatically.

Example: Selecting frame 30 to set the (animated) camera viewpoint to some actual picture which was shot in portrait mode and inserted in frame 30 of the background video:

The 30^rd background video frame is stretched because the "resolution" is not set accordingly. Is there a bypass?

Answer 1

This is my solution, greatly inspired from @sambler elements (please give him credits, not me).

import bpy

# Define size elements
mx, mn = 1500,1000

# There are less portrait frames, let see them as exceptions
portrait_frames = 11,12,13,14,16,19,24,27,30,31, \
                  33,34,35,36,38,39,40,42,43,44,45

# Define a callback function to call when scene current frame changes
def chg_res(scene):
    # Get frame number
    f = scene.frame_current

    # Set desired resolution for this frame
    if f in portrait_frames:
        scene.render.resolution_x = mn
        scene.render.resolution_y = mx
    else:
        scene.render.resolution_x = mx
        scene.render.resolution_y = mn

# Register callback in Blender event system (once)
bpy.app.handlers.frame_change_pre.append(chg_res)

and the result:

(The frame is manually changed. This leads to the camera resolution being changed by the callback function. The background image and the camera position/orientation are changed by normal animation keyframing).

Answer 2

Thanks to some help from Astronet and Sybren, I have code that works for rendering. Hope this helps anyone who was stuck like me.

    import bpy
    
    res_1 = [2560,941]
    res_2 = [1700,540]
    res_3 = [313,550]
    
    
    frame_res = {
        1: res_1,
        2: res_2,
        3: res_3,
        }
        
scene = bpy.context.scene
render = scene.render
f_path = render.filepath

for frame, resolution in frame_res.items():
    scene.frame_set(frame)

    render.resolution_x, render.resolution_y = resolution
    render.filepath = f'{f_path}{frame}'
    
    bpy.ops.render.render(write_still=True)
render.filepath = f_path

Answer 3

I think the only way to do that is to use a python script to adjust and render each frame.

import bpy

res_landscape = [1920,1080]
res_portrait = [1080,1920]

frame_res = {
    1: res_landscape,
    2: res_landscape,
    3: res_portrait,
    4: res_portrait,
    5: res_landscape,
    }

scn = bpy.context.scene

for f,r in frame_res.items():
    scn.frame_set(f)
    scn.render.resolution_x = r[0]
    scn.render.resolution_y = r[1]
    scn.render.filepath = '//renders/image{:03}.jpg'.format(f)
    bpy.ops.render.render('INVOKE_DEFAULT', write_still=True)