The way I achieved this was faking orthographic view in perspective mode, by using the “Calculating distance” formula on [this wikipedia article about Dolly Zoom][1]. Basically, pull away your camera a whole lot and then zoom in tight.
I was using a CineCamera object on a Camera Boom.
For perspective, I manually set a good boom length and focal length (which controls the FOV).
Noted these values down and plugged into the formula to get the initial “width” (which may be different from your screen width).
Now for orthographic, I set a big value for boom length (eg. 80000) and plugged the above calculated width to get the FOV value, and set it into the camera.
(In this example, for perspective, I pass Boom Len as 1500, and for orthographic, I pass the earlier mentioned 80000)
For a smooth transition, you can hook Set Camera Projection Params up into a timeline.