Text regions of natural language embedded in images are automatically translated via a clipboard buffer during a cut-and-paste command according to an isolation rule defining one or more types of image elements to be isolated, by examining the image elements, responsive to meeting the isolation rule, transferring the image element from the clipboard to an isolation store, disabling pasting of the image element; recognizing text within the isolated image; determining a natural language in which the recognized text is expressed; performing natural language translation of the recognized text into a user-specified destination natural language; producing image data for the translated text; overlaying the image data onto the embedded text in the isolated image element; and pasting the modified image element from the isolation store to a user-designated destination points within a destination computer file.