The objective of this study is to present a database of metadata, which allows effective research into newsreels. Old sound newsreels have great potential for research to learn more about how people spoke many decades ago. Historians and media researchers have highlighted the significance of voice-overs on newsreels from their professional point of view. However, only few studies are known which used methodology of linguistic research. Since the speaking material of the newsreels is highly diverse, and the identification of the utterances relevant to a specific research project is crucial, a database of metadata is of importance. Several suggestions of previous authors are presented. The second part of the study discusses a possible structure and the use of a relational database, which stores metadata for old Hungarian newsreels. This database is under development, but it is expected that when it is ready, it will help researchers efficiently find the recordings they need. Examples of possible uses are included, such as the identification of utterances belonging to a specific speaking style, analysis of within-speaker variability, or improvement of intelligibility of noisy or distorted recordings.