The recent and fast development of automatic transcription software is accompanied by a growing heterogeneity of formats to save the output of such a task. TEI P5 can be helpful to simplify workflows and bring in more coherence in digitization pipelines. We present a twofold modelization in TEI which brings together essential information resulting from the transcription phase with the editorial layers. The usefulness of this modelization is illustrated with several examples showing how such an approach can be leveraged at different stages of a digitization pipeline.