This paper analyses the steps in the writing of the book Variations sur la linguistique, in which Culioli is interviewed by Fau. First appears the dual constraint of that writing: from the oral speech recorded on digital file, the writer has to elaborate a written speech, i.e. governed by rules of structuration and organisation of the written language, but at the same time the speech must use oral markers. The textual genetic analysis of the four successive versions of draft documents, from the first oral-into-written transcription to the printed version, allows a chronological description of that dual work. By using the meta-enunciative categories proposed by Authier-Revuz in her works on the representation of other’s discourse, the analyses determines any discernable patterns of this specific situation, in which each stage consists of a reformulation of the precedent version and at the same time a representation of the original oral speech.