Technology is disclosed for enabling scene-based variable compression (“the technology”). In some embodiments, the technology can receive an indication of a compression level for a first content type, wherein the compression level specifies a tolerance level for lossy compression; receive a request for content, the content having at least two portions, wherein a first portion has first content of the first content type and the second portion has second content of a second content type, a first compression method associated with the indicated compression level and a second compression method associated with a different compression level. The technology can transmit the content, wherein the first portion of the content was compressed using the first compression method and the second portion was compressed using the second compression method.