Diff Anything chooses a comparison engine that understands the inputs. Text uses a focused side-by-side diff, JSON and other structured formats compare semantic paths, CSV can match rows by key, folders recurse with ignore rules, and images add pixel heatmaps, overlay, and blink views. Compared files never leave the computer. There are no accounts, cloud comparison services, analytics, or telemetry. CLI and Git difftool modes make the same comparison model available in scripts and source-control workflows.