We present an open-source speech corpus for the Kazakh language. The Kazakh\nspeech corpus (KSC) contains around 332 hours of transcribed audio comprising\nover 153,000 utterances spoken by participants from different regions and age\ngroups, as well as both genders. It was carefully inspected by native Kazakh\nspeakers to ensure high quality. The KSC is the largest publicly available\ndatabase developed to advance various Kazakh speech and language processing\napplications. In this paper, we first describe the data collection and\npreprocessing procedures followed by a description of the database\nspecifications. We also share our experience and challenges faced during the\ndatabase construction, which might benefit other researchers planning to build\na speech corpus for a low-resource language. To demonstrate the reliability of\nthe database, we performed preliminary speech recognition experiments. The\nexperimental results imply that the quality of audio and transcripts is\npromising (2.8% character error rate and 8.7% word error rate on the test set).\nTo enable experiment reproducibility and ease the corpus usage, we also\nreleased an ESPnet recipe for our speech recognition models.\n