Since this is an AI-based process, some fine detail is reconstructed by the model rather than recovered from the source. For this reason, on-screen text (signs, labels, credits) and background details may appear distorted or differ from the original.
