The objective of this study was to test the performance of a prototype vision system in phenotypically diverse beef and lamb carcasses against visual grading of eye muscle area (EMA), marbling and chemical intramuscular fat (IMF%). Validation in beef demonstrated that the camera prototype in combination with analytical techniques enabled prediction of EMA (r = 0.83, RMSEP = 6.4 cm), MSA marbling (r = 0.76, RMSEP = 66.1), AUS-MEAT marbling (r = 0.70, RMSEP = 0.74) and chemical IMF% (r = 0.78, RMSEP = 1.85%). Accuracy was also maintained on validation with all four traits displaying minimal bias of -3.6, 6.3, 0.07 and - 0.01, for EMA, MSA marbling, AUS-MEAT marbling and IMF% respectively. Preliminary analysis in lamb indicates potential of the system for the prediction of EMA (r = 0.41, RMSEP = 1.87) and IMF% (r = 0.28, RMSEP = 1.10), however further work to standardise image acquisition and environmental conditions is required.