Systems and methods for dynamically and automatically generating short-form versions of long-form media content are provided. The long-form media content may be tagged with metadata indicating objects or actions in frames, scenes, portions of scenes, or other units of the media content. The systems and methods may receive a user-specified time limit and use the metadata to create a short-form version of the long-form media content (using one or more of the scenes, portions of scenes, or other units of the media content), which preserves one or more story arcs within the media content.