Abstract. Completing missing information of signals was for a long time solved by copying samples from different parts of the signals or using several interpolation techniques. Recent research in sparse representations of signals provides a new approach for completing missing samples called Audio Inpainting. In the talk the basic idea of Audio Inpainting will be presented, including some approaches for specific applications. A brief introduction to dictionary learning and its connections to Audio Inpating will be given. Current work using the Constant-Q transform based on Nonstationary Gabor Frames will be described as well as future ideas.