Google patents virtual meeting views from any angle

1. A method comprising: obtaining, by a processing device associated with a platform, first image data generated by a first client device at a particular location during a virtual meeting and second image data generated by a second client device at the particular location during the virtual meeting, wherein the first image data depicts one or more objects captured from a first vantage point of the particular location and the second image data depicts the one or more objects captured from a second vantage point of the particular location; receiving, by the processing device, a request from a third client device for third image data depicting the one or more objects captured from a third vantage point, wherein the first vantage point and the second vantage point surround the third vantage point; determining whether the third image data depicting the one or more objects from the third vantage point is to be generated; responsive to determining that the third image data is to be generated: providing the first image data, the second image data, and an indication of the third vantage point as input to a machine learning model, wherein the machine learning model is trained to predict characteristics of an image depicting objects captured from a particular vantage point based on characteristics of given image data corresponding to two or more different vantage points; and extracting, from one or more outputs of the machine learning model, characteristics of the image depicting the one or more objects corresponding to the third vantage point; generating, by the processing device, the third image data depicting the one or more objects corresponding to the third vantage point based on the obtained extracted characteristics; and providing, by the processing device, a rendering of the third image data for presentation via a graphical user interface (GUI) of the third client device during the virtual meeting in accordance with the request.