This dataset consists of 100 hours of conversational speech segmented by speaker and transcribed for training speech recognition models. It includes various languages such as English, Bulgarian, German, Lithuanian, and Norwegian, with each recording segmented into speech, noise, and music.