UV-chemiluminescence from the excited hydroxyl-radical (OH*) has been used as a marker for the high-temperature reacting zone in spark-ignited engines for quite some time. In research engines with large optical access, sensitive camera systems make it possible to obtain images of the flame that can be used for, e.g., determining the flame-front's propagation speed [Aleiferis et al., Combust. Flame 136 (2004) 283-302]. However, on one hand such optical engines are limited in their speed and load range, on the other, typical UV endoscopes make wide-field imaging at low light levels challenging.Here, a large-aperture UV endoscope is used to capture sequences of OH* chemiluminescence during early flame propagation in a nearly unmodified production engine. We compare three imaging systems: phase-locked single-shot imaging, phase-locked double-frame imaging, and “high-speed” cinematography at kHz repetition rates. The four-cylinder spark-ignition engine can be operated at speeds and loads significantly exceeding the limits of most fully optically accessible engines. During the first 20° crank-angle after ignition, the phase-locked endoscopic images almost match the image quality reported from experiments in a dedicated optically-accessible engine. For later acquisition timings, the flame often exceeds the field of view. From phase-locked imaging the instantaneous size of the apparent burnt area (ABA) can be identified by thresholding after filtering. Its single-shot variant allows only computation of the multi-cycle average of the apparent flame speed (AFS). Acquisition of two successive frames in a single cycle enables determining the instantaneous AFS. High-speed imaging can follow a single cycle and thus the time-resolved ABA can be estimated, but the instantaneous shape of the flame cannot be imaged with much detail, because the detector hardware is less mature.