Semantic Transfer of ISelective-fidelity semantic image transfer aims to preserve task-relevant content at high quality while conveying only sparse evidence from the remaining regions, delegating completion to a generative decoder. Under strict rate constraints, two coupled transmitter decisions become critical: selecting the Region of Interest (ROI) among multiple plausible segments and sampling a limited set of patches over the Region of Non-Interest (RONI) that anchor generative inpainting. In this work, a state-of-the-art selective-fidelity semantic image transfer framework called SPIFF is considered. An encoder-side instantiation is developed within a SPIFF-like pipeline: the receiver-side diffusion inpainting module is kept fixed, and the study intervenes exclusively on transmitter-side decisions and policies.