GhostTeX 3.0 adds a lower-memory local AI option for Apple silicon Macs with 8 GB of unified memory. • New Qwen3.5 Light engine: a 2-billion-parameter, 8-bit model with a 2.7 GB download. Choose Light for clean, typed pages on an 8 GB Mac; Fast and Precise remain available for handwriting and complex layouts on Macs with more memory. • Safer Fix Title, Authors & Abstract: GhostTeX now checks the AI result against the source, preserves author affiliations, and prevents incomplete AI responses from erasing or shortening abstracts and keywords. • More reliable LaTeX conversion: improved section and reference detection, Greek notation, author markers, bibliography ordering, and handling of special characters that could previously break a build. • A cleaner workspace: LaTeX syntax highlighting, easier access to TeX Diagnostics and build controls, smoother streaming, and cancellable model downloads. AI processing remains on your Mac. After the one-time model download, your documents are not sent to an AI service.

A cleaner, redesigned GhostTeX: • One AI Engine control with Fast and Precise modes • Browser-style Full Document / Equation Capture tabs, and support for light and dark appearance that follows macOS • New "Fix Title, Authors & Abstract" button: GhostTeX finds the real title page (it isn't always page 1), shows what it found, and lets you scan a different page • Clearer TeX project folder and Build PDF controls • Larger chat box: Return sends, Shift+Return adds a new line • More reliable AI document analysis and free AI credit checks

Try AI features with 10 free credits • Convert only the PDF pages you choose • Credits are charged only when an operation succeeds • Subscribers continue to receive unlimited AI usage • Improved reliability, cancellation handling, and quota synchronization • Additional stability improvements and interface refinements

GhostTeX 2.0.1 — a major update focused on speed, reliability, and control. What's New • Scan straight from your iPhone or iPad. Take a photo, scan a document, or add a sketch on your phone and it appears in GhostTeX ready to convert — no cables, no AirDrop, no saving to Files first. Photograph a page from a book, a printed handout, or a whiteboard and go straight to LaTeX or Markdown. (Requires an iPhone or iPad signed in to the same Apple Account, with Bluetooth on.) • A new Settings window (⌘,) with a Dark Mode toggle and a full Models pane: download, load, delete, and relocate your AI models in one place. • See your source and the compiled PDF side by side. No more switching back and forth to check your work — edit on one side, watch the PDF update on the other after each build. • Autosave and recovery. GhostTeX now quietly saves your work in the background. If the app quits unexpectedly, you'll be offered to restore right where you left off. • A new, faster AI model option. Alongside the existing model, you can now choose a lighter, quicker one for straightforward documents where speed matters more than handling complex layouts. • Chat and document conversion can no longer interfere with each other — each waits its turn instead of racing. Plus numerous stability and reliability fixes under the hood.

GhostTeX now supports session save and restore. You can save your current workspace, including the source PDF or image, extracted LaTeX/Markdown, build state, and chat context, then reopen it later and continue where you left off. Switching between Full Document and Equation Capture also preserves your work in each mode.

• Broader reference/bibliography support (bracketed, numbered, and heading-less lists) • Fixed caption compile errors (math, special symbols, unbalanced $) • Fewer LaTeX build failures on real papers

• Line numbers in the editor, matching the line references in build errors so you can jump straight to a problem. • More reliable text display — converted and streamed text now renders correctly as pages come in. • Redesigned TeX Diagnostics window: clearer compiler errors and pre-build source checks, each with a one-click "Go to Line." • Smarter Markdown-to-LaTeX conversion — cleaner handling of math and subscripts for fewer compile errors. • Safer undo/redo and various stability fixes.

What’s New in GhostTeX 1.2.1 Added optional AI Refinement to reconstruct a paper’s title, authors, affiliations, abstract, and document structure. AI refinement uses evidence from the first two pages and avoids inventing unsupported metadata. Selected-page text is preserved during refinement. Added local LaTeX compilation with live PDF preview. Improved TeX error reporting with source-line context. Build logs can now be selected and copied. Improved Markdown export for Typora and Obsidian. Improved stability and performance.

What’s New in GhostTeX 1.1.1 GhostTeX now uses an updated MLX engine for faster, smoother on-device AI performance when converting PDFs, scanned pages, and handwritten equations into editable LaTeX. This release improves the local model runtime on Apple Silicon Macs while keeping the same private workflow: your documents stay on your Mac, with no cloud upload or API key required.

Subscription is fixed.