As the name suggests, all the rendering is done up front. The camera positions are predefined, and the generated visuals are stored as images. Or you can create video by rendering a sequence of images and moving the camera or the object. You can also use this technique to generate all the views around the object to create a turntable effect. The end-user can interact with the image sequence and display their chosen view of the object, if it is one of the pre-rendered views on the turntable. (This is sometimes referred to as 2.5D)
Pre-rendering also means you can choose the software used to create the render, allowing for advanced visual creation of textures like hair, or movement of cloth over an arm.
With pre-rendering, the speed of rendering does not matter. It can take hours, if required, to render a single image to allow for huge amounts of detail, realistic textures, and real-world quality. It also does not matter what computer the end user is using to view the rendered image, as all the effort is done upfront. The end user only see the results: the final image or the recorded video. The big drawback with pre-rendering is that the user cannot change what they see, they are limited to the decisions made at the time the rendering took place.