Developers often refer to video-hosting online platforms to find screencasts that provide a step-by-step guide to help them solve a programming task at hand or learn a new concept. More specifically, developers search for resources that help them design and implement effective mobile graphical user interfaces (GUI) using XML. Although mobile programming screencasts contain a vast amount of XML data at developers’ disposal, they cannot be easily found and copied-pasted due to the image nature of videos. Given that the most common task developers perform online is copy-pasting, mobile programming screencasts must support that and be complemented with XML data in a textual format. To overcome this challenge and aid developers, this paper presents vid2XML, which is a three-phase approach that leverages both visual and textual information of video frames to locate XML region in video frames, locate the currently opened file, and extract XML data for each file presented in video frames. We evaluated each phase of vid2XML in a comprehensive empirical evaluation on videos collected from YouTube. The results reveal that vid2XML is able to accurately (i) locate XML regions, outperforming four previous work, (ii) locate the bounding box of the selected file, and (iii) extract, fix, and merge XML data for each file opened/created in a video.