In this talk we deal with Information Retrieval from audio-visual recordings. Such recordings are often long and a user may want to know the exact relevant passage of the recording. Therefore, the recordings are automatically divided into smaller parts, on which we apply standard retrieval techniques. We experiment with various methods for segmentation of the recordings into shorter segments which are then used in a standard retrieval setup to search for relevant passages. The main focus will be on machine-learning based approaches utilizing the content of the recordings. We will describe experiments performed in two shared tasks of MediaEval Benchmark: Similar Segments in Social Speech and Search and Hyperlinking.