This paper offers a multidisciplinary reflection on filmic practices of researchers in anthropology and cognitive ergonomics, within the framework of their understanding of rural knowledge confronted with the question of development. The video seems to be a support as well as a medium: it permits the object of “know-how” to be tackled as a complex construction, consisting of interactions between technical, relational and cognitive components. It also facilitates new interactions between the actors, whether they “hold” or “promote” some know-how, when they are involved, for example, in the same analytical and prospective step in the technical performance, or the cultural and political dimensions of this “know-how”. Using methods of self-confrontation or crossed confrontation, the image recorded allows the operator to reconsider his activity and clarify his gestures, strategies, the environmental indicators taken into account within the framework of certain decisions or adaptations etc. The image is, in this case, a tool for formalising know-how. The work represented, rebuilt by the image, is also rebuilt by editing, a strategic moment of reinterpretation and analysis. Last but not least,the video sequence, according to the researcher’s ethics (which should be analyzed), plays a clear role of mediations “catalyst” between the protagonists and as a particular medium, makes it possible to modify a context of negotiation or to build a consensus on united actions. Restitution becomes a mediating representation for professional worlds, sometimes distant from one another, which develop through it a consensus, the sharpening of a utopia or a project. This paper, based on various texts and field experiments, will specify these dimensions of the video and the image. It will then identify certain limits to these tools/supports/mediators in our research practices and development support: from the cultural gaps between the observers and the observed, the obvious risks of over-interpretation, to the media abuses of the ‘film direction’, use of video necessitates vigilance at various stages of the research and collective action if one wants simultaneously to produce an interaction, to film the action and to scientifically validate the data converted into images.