This paper deals with an adaptation of multimodal transcription as devised by Paul Thibault and Anthony Baldry (Baldry ed. 2000). This adaptation is designed to provide a basis for a reasoned approach to the subtitling of video products, particularly cinema and television. The system has already been adopted in the analysis of various different film types. The process involved has enabled the author to devise a minimalist subtitling strategy based on the interplay of all the semiotic modalities contained in each video. Secondly the array of film types chosen allows for a comparison of the strategies adopted, which can change from one category to another (e.g. the unseen monologues in documentaries may require a different approach to the animated dialogues in soap operas). Finally, the process of creating a multimodal transcription consisting of detailed annotations of single frames of film is in itself a useful didactic exercise.