Diff Anything 1.5 adds Smart Diff as an optional code view while keeping the conventional Line Diff available.

Smart Diff aligns language-aware tokens through formatting and reflow. Line Diff remains the literal line-by-line view. Both retain normal syntax highlighting, exact source text, editing, moved-block cues, and reviewed apply actions. Smart Diff uses conservative language-family rules rather than claiming compiler-level AST equivalence.

The view is available for all 117 non-plain-text editor languages. Version 1.5 adds maintained definitions for Ada, Assembly, COBOL, Crystal, CUDA, D, Elm, Erlang, Fortran, GLSL, Haskell, Haxe, HLSL, LaTeX, LLVM IR, OCaml, Raku, and WebAssembly text.

This release also improves large-comparison performance, cancellation, virtualization, interaction feedback, accessibility, and recovery. Everything runs locally on macOS and Windows, without accounts, cloud comparison, analytics, or telemetry.

Diff Anything 1.4 is out, with a much broader archive comparison engine.

It now compares ZIP/ZIP64/ZIPX, ZIP-based packages, 7z, RAR, TAR families, standalone Gzip, Bzip2, XZ, and Zstandard streams, encrypted ZIP/7z/RAR archives, and supported multipart 7z/RAR sets. Entries are matched by normalized path and streamed content hash, so timestamps, compression levels, member order, and other packaging-only changes do not create false differences.

The three-way merge workspace has also been redesigned. Ours and Theirs are each compared with Base, conflict navigation is clearer, unchanged source lines can collapse, bulk Ours/Theirs choices and reset are available, and compatible character-level changes on the same line can merge automatically.

Everything runs locally on macOS and Windows. There is no account, upload, cloud comparison service, analytics, or telemetry.

Diff Anything 1.3 is out, and cross-format spreadsheet comparison is the biggest change.

You can now compare stored cell values across XLSX, ODS, Apple Numbers, CSV, and TSV even when the two files use different formats without converting first. Spreadsheet mode deliberately compares stored values, not formulas, formatting, charts, macros, or workbook layout.

I also rebuilt folder diff around two-pane tree and table views, search, status filters, moved-file detection, and safer synchronization plans. PDF review now separates visual page comparison which can reveal changed charts, images, visible signatures, stamps, and layout from extracted-text comparison for wording changes.

Everything still runs locally on macOS and Windows. There is no account, upload, cloud comparison service, analytics, or telemetry.

I built Diff Anything because I kept running into comparison tools that worked for one input type but broke down when I moved between text, JSON, folders, images, and PDFs.

Version 1.2 is mainly about making those difficult cases more predictable. Ambiguous structured inputs now fail more clearly, repeated headers and keys are handled more carefully, masked environment values stay masked, image transparency and large decoded images have safer boundaries, and PDF text is kept separate from rendered-page evidence.

The desktop app, CLI, Git integration, and batch workflows now share the same comparison boundaries. You can try every desktop comparison engine for free within explicit preview limits; the free CLI remains text and JSON.

Everything runs locally there's no account, cloud comparison service, analytics, or telemetry.