Textometry is applied to audiovisual corpora, such as transcripts from semi-directed interviews, or the "Actualités françaises" newsreels archive. A workflow using an assisted or automatic transcription software is efficient to get a rich encoding. New features are added to the TXM software: a specialized import module based on Transcriber XML format, a utility to convert text transcripts to XML, and the MediaPlayer extension to watch the video segment corresponding to a word context selection. Methodological thoughts arise from this experience. It is highly relevant that textometry takes into account internal text structures (such as speech turns) and other meta-information (such as timecodes). Meta-information has to be displayed and available for processing without being mixed with contents. Another challenge is to integrate multiple interrelated representations. A back-to-media feature is as fundamental as the back-to-text one to provide context to interpretation work.