Darragh O'Brien's main research interest is speech processing, with a particular emphasis on concatenative speech synthesis.

In concatenative synthesis, the prosody (pitch, duration and intensity) of component speech units, retrieved from an inventory at synthesis time, will usually not match the target prosodic context. Units must therefore undergo significant modification. Furthermore, smoothing of spectral and phase discontinuities across unit boundaries will be required. Thus, the quality of a speech synthesis system's output is crucially dependent on the efficacy of both its underlying smoothing and pitch and time-scale modification techniques. Dr. O'Brien has conducted research into, and continues to be interested in, the area of prosodic modification.

In an effort to reduce unit smoothing requirements, unit-selection-based synthesis systems have been developed. Employing large inventories (containing multiple instances of any given synthesis unit), such systems have been shown capable of producing very high quality, natural sounding speech output. During synthesis, an utterance is built by concatenating an optimal sequence of speech units extracted from the inventory. To produce natural sounding speech, an attempt is made to choose that sequence which minimises the perceived level of discontinuity across the resulting output. In order to minimise it, it is necessary to be able to measure perceived discontinuity. Dr. O'Brien's current research centres on developing perceptually salient distance metrics for this task.