This paper presents a framework for simulating visually realistic motion of underwater Remotely Operated Vehicles (ROVs) in highly complex models of aquatic environments. The models include a wide range of objects such as rocks, fish and marine plankton in addition to an ROV tether. A modified cable simulation for the underwater physical conditions has been developed for a tethered ROV. The simulation framework also incorporates models for low visibility conditions and intrinsic camera effects unique to the underwater environment. The visual models were implemented using the Unreal Engine 4 realistic game engine to be part of the presented framework. We developed a generalized method for implementing an ROV dynamics model and this method serves as a highly configurable component inside our framework. In this paper, we explore the unique characteristics of underwater simulation and the specialized models we developed for that environment. We use computer vision algorithms for feature extraction and feature tracking as a probe for comparing experiments done in our simulated environment against real underwater experiments. The experimental results presented in this paper successfully demonstrate the contribution of this realistic simulation framework to the understanding, analysis and development of computer vision and control algorithms to be used in today’s ROVs.