This report describes LIMSI's submissions to the Round 1 of translation shared task at MLIA. We have submitted systems using various combinations of in-domain and out-of-domain corpora fine tuned on the MLIA COVID data, in order to develop resource-heavy systems for the translation of COVID mlia data from English into French, using two pre-processing pipelines and pretrained representations. Our experiments demonstrate the effectiveness of in-domain and task specific corpus, even if in small amounts.