You (don't) simply press a button to transcribe an interview!

We present a new IT service at Friedrich Schiller University Jena, that allows researchers to transcribe audio recordings in high-quality and in a privacy-friendly manner. The service seamlessly integrates in the university’s Nextcloud, where researchers can upload their recordings and transcribe them one by one or in bulk. All components are entirely based on open-source software and open machine-learning models and the service processes the recordings exclusively within the university network on university-owned hardware. While Nextcloud already offers an interface for AI functions, its current implementation only connects to applications run locally on the Nextcloud server. Exploiting a new language-agnostic AppAPI plugin, our service connects users transparently to the university HPC cluster, separating storage and compute. We present two perspectives of this project. Firstly, we present how the concept for this service emerged and evolved, and describe the technical solution we ended up realizing. Secondly, we use this example as a case study to show of how RSEs can foster collaboration between players in universities and lubricate the research process, in this case, by providing automated workflows.