The COVID-19 pandemic and associated restrictions on in-person collection created a need for researchers to familiarise themselves with online data collection methods, often through platforms that allow for participation in the absence of a researcher. The development of online methods could prove advantageous by enhancing access to musical populations. Recent psycholinguistic studies employing web-based audiovisual tasks have reproduced lab-based findings with good data quality. However, the online collection of musical data has not been widely explored. Here we offer a tutorial to introduce the first use of the Gorilla Experiment Builder to collect musical improvisations played in response to prompt and recorded on participants own instruments at home in their own time. We aim to support other researchers in developing similar tasks. To test the method, we recruited 76 experienced pianists through Prolific who completed an improvisation task that allowed for participants to test their own recording quality, included written and video instructions, allowed them to hear a musical prompt and metronome, and view the associated musical score on the same screen. Technical issues faced included incomplete improvisations and poor sound quality, but included participants produced high quality data, particularly after the inclusion of an explanatory video. We propose that data collection using online and self-paced task could prove a fruitful vehicle for certain types of musical research and potentially greatly reduce the resources required to collect data.