This talk deals with the challenges of remote vs. on-site fieldwork and of the documentation/description of an almost extinct language: Kibiri (isolate) a highly threatened and poorly known language (Hammarström 2017 : 297; Okura 2017 : 351, 360; Pawley & Hammarström 2018: 81) spoken in some tributaries of the Kikori river in the Gulf province of Papua New Guinea. Due to the Covid-19 pandemic, access to the field site was impossible; therefore I carried out remote fieldwork for 5 months with a single speaker in 2021. I am currently carrying out a long on-site fieldwork of 10 months until late 2022 on this language, whose actual number of speakers is estimated at around 30 (direct observation, unlike Ethnologue numbers of 2011: 280). The main reasons of language loss are reported by the Kibiri themselves to be interethnic marriage with non-Kibiri speakers for the last 50 years, interdiction of speaking the language by Papuan missionaries in the 1960’s and the subsequent lack of intergenerational transmission. This talk will discuss the opportunities and limits of both fieldwork modes based on my experiences as a fieldworker and the speakers’, in order to shed light on the (im)possibilities of documenting and describing severely endangered languages in the best and most complete way possible. Some observations that can be advanced are that there are many factors influencing a remote fieldwork situation (Lesage et al. forthcoming) and that the lack of language use has a negative impact on the documentation of the language in its social and cultural context, thus making its description less diverse.