Model tier override. Every AI transform normally picks its own model tier — Light, Standard, or Advanced — and the new tier control in the full window's top bar (next to Settings) lets you take that decision over: leave it on Auto for the old behavior, or lock every AI transform to one tier while you compare models. The menu lists the model each tier currently resolves to on your provider, so it's obvious what you're switching between. The setting is global and sticks across restarts, and the command palette honors it too — while a tier is locked, the control fills in so you can't forget it's on. In Auto the control also names the tier the highlighted transform will use ("Auto · Standard").
Transform timing. The full window now reports how long each AI transform took, under the output box next to the character counts: elapsed time, the tier it ran at, and the exact model — e.g. 2.41s · Standard · gemini-3.5-flash. The timer counts up live while the result streams and freezes the moment streaming ends, so slow models announce themselves early. Handy for comparing local models against each other, or against a hosted one.
Rewire Text no longer fails on Windows 10 the first time it uses its database. Transforms could die with Invalid argument(s): Failed to lookup symbol 'sqlite3_stmt_isexplain' ... (error code: 127). The app now ships its own copy of SQLite instead of quietly relying on the one built into Windows, which on Windows 10 is too old to work. Windows 11 was not affected.
Per-transform hotkeys. Give any transform its own global hotkey: press it in any app and Rewire Text captures the selected text and opens the command palette with that transform already selected and running, so you can review the preview and press Enter to apply. Set them up in the new Settings › Hotkeys page — add, edit, or remove hotkeys, with conflicts against the global hotkey or other assignments caught as you record.
Skip preview (YOLO). When assigning a transform hotkey, you can opt to apply the result immediately instead of stopping at the preview. Instant (non-AI) transforms paste straight back without opening the palette at all; AI transforms briefly show the palette with a "YOLO — applies when done" badge while the result streams in, then paste back automatically (errors still keep the palette open so you can see what happened).
Vim mode. Turn it on in Settings › General › Keyboard to drive the command palette (and the full window) from the home row: j/k move through the transform list, i jumps into the search box to filter, Enter applies, and Esc steps back — out of the search box first, then closing the palette. In the full window, h/l collapse/expand categories. Off by default; when off, everything works exactly as before.
Keyboard navigation in the full window. Even with Vim mode off, the transform list now responds to the arrow keys once it has focus: ↑/↓ move the highlight, ←/→ collapse and expand categories, and Enter applies the highlighted transform (or toggles a category).
Changed
The global hotkey setting moved from Settings › General to the new Settings › Hotkeys page, so all hotkey configuration lives in one place.
Fixed
Recording a hotkey now works the natural way — hold the modifiers, then press the key. Before, the recorded key would flicker and vanish (a held modifier's key-repeat overwrote it), and only pressing all keys simultaneously would sometimes stick. The recorder is now a click-to-record field: held modifiers show as a live preview, the chord locks in when you press the main key, and it no longer picks up keys typed into other fields of the same dialog. While recording, the app's registered hotkeys are paused so you can re-record the combination you're already using without triggering it.
Automatic update checks. Rewire Text now checks for a new version in the background — about once a day — and shows a slim notice at the top of the window (and a highlighted item in the tray/menu-bar menu) when one is available. It's quiet by design: nothing pops up unless there's actually an update. From the notice you can open the changelog to download, skip that particular version, or dismiss it for now. You can turn automatic checks off entirely in Settings › General › Updates; the manual Check for Updates action is always available regardless.
Custom transform presets. Save your own AI instruction as a reusable transform: give it a name and it shows up in the list under Custom as "Custom: <name>", ready to run like any built-in transform — no need to retype the instruction each time. Create, edit, and delete presets in Settings › Transforms, and pick which model tier (Light / Standard / Advanced) each one runs at. Type "custom" in the search box to see all your presets alongside the built-in Custom Instruction.
Chain transforms in the command palette. After a result previews, press Tab to feed it back in as the input for the next transform — run as many in a row as you like without leaving the palette. A breadcrumb at the top shows the chain you've built, and a button on the preview does the same as Tab for the mouse. Press Enter on the final preview to accept it, or ⌘⏎ / Ctrl+Enter to accept the chained result at any point (even between steps).
Undo/redo while chaining. Step back and forth through the chain with the standard shortcuts (⌘Z / ⇧⌘Z on macOS, Ctrl+Z / Ctrl+Y on Windows) — or click any step in the breadcrumb to jump straight to it. Undone steps stay visible (ghosted) until you build past them. Pressing undo while a result is previewing simply discards that preview.
Anonymous usage check-in. The app now sends an anonymous ping on launch — at most once a day — so we can count installs and see how many trials convert to a purchase. It includes only a random per-install id, the app version, your OS, whether a license is active, and (once licensed) a one-way fingerprint of your license key so your own devices aren't counted as separate customers. It never sends your name, email, license key, or any of your text.
Changed
Settings reorganized. The old "AI Transforms" page is now "AI Models" (provider, API keys, and per-tier model selection). Translation languages have moved to a new "Transforms" page, which is also where your custom transform presets live.
In the command palette, ⌘⏎ / Ctrl+Enter now accepts the result (see above) rather than opening the full window. You can still open the full window from the palette with the expand button in its top-right corner, or by pressing your global hotkey a second time while the palette is open.
Keyboard-shortcut hints throughout the palette (footer, tooltips) now show the correct keys for your platform — ⌘-style glyphs on macOS, Ctrl+… on Windows — instead of a combined label.
Title Case now follows headline-style rules: articles (a, an, the), coordinating conjunctions (and, but, or, nor), short prepositions (to, in, by, of, for, on, at, off, up, via, per), and "as" are kept lowercase when they fall in the middle of the title — but are still capitalized as the first or last word. So "the lord of the rings" becomes "The Lord of the Rings".
Fixed
Slugify now drops apostrophes instead of turning them into hyphens, so contractions and possessives stay whole — "Peter's new computer" becomes peters-new-computer rather than peter-s-new-computer.
Title Case now lowercases ALL-CAPS input before capitalizing each word, so "THIS IS JUST A TEST." becomes "This Is Just a Test." instead of being left shouting. (Acronyms in all-caps text are lowercased too, e.g. "NASA" → "Nasa".)
The command palette's result preview is now resizable — drag the handle along its top edge to make the preview taller and read more of a long result before you accept it.
xAI (Grok) is now a built-in AI provider. Pick it in AI settings and use Grok models across the Light / Standard / Advanced tiers (Grok 4.20 / 4.3 / 4.5), with cost tracking like the other providers. Requires an xAI API key.
The model picker in AI settings now shows each model's price (input / output per million tokens) on the right of the row, so you can weigh cost against capability when choosing a model per tier.
Changed
Usage & cost tracking now estimates costs from an up-to-date pricing catalog covering every major provider, so far more models show a real cost — and prices refresh automatically instead of being frozen at each app release. OpenRouter models are priced from the live rates OpenRouter itself reports.
The recommended Anthropic "Advanced" model is now Claude Opus 4.8 (previously Claude Opus 4).
Fixed
The command palette now remembers its size. Resize the palette window and it reopens at that size the next time you summon it, instead of snapping back to the small default.
Corrected some out-of-date model prices in the usage summary (for example, Claude Haiku 4.5 is priced at $1.00 / $5.00 per million input / output tokens).
When an update is available, the update dialog now opens the online changelog — where you can see everything that's new since your version and download the update — instead of downloading the installer directly.
Jump from the command palette to the full window without losing your place: click the expand icon in the palette's top-right corner, press Ctrl/⌘+Enter, or press your global hotkey a second time while the palette is open. Your captured text lands in the input box and the transform you had highlighted stays selected.
The Windows installer now installs just for you (into your user folder) and no longer prompts for administrator approval. Existing installs under Program Files aren't removed automatically — uninstall the old copy once if you'd like to tidy up.
Chain transforms in the full window: a new "Use as input" button on the Output pane feeds the result back into the Input pane, so you can run several transforms in sequence without copying and pasting.
Launch at login: start Rewire Text automatically when you log in (Windows and macOS 13+). When started this way it opens quietly in the menu bar / tray instead of showing a window. Toggle it in Settings › General or from the tray menu; a one-time prompt offers to turn it on for you.
Global-hotkey command palette + full-window mode for transforming selected text in any app.
43 deterministic transforms — case conversion, formatting & cleanup, encoding/decoding, developer tools, and Markdown.
BYOK AI transforms (tone & style, writing assistance, structure, translation, custom) across OpenAI, Anthropic, Gemini, OpenRouter, and local Ollama, LM Studio, and Llama.cpp with streaming results.
API usage and estimated-cost tracking.
7-day full-featured trial with Creem licensing; a free-forever subset of deterministic transforms remains usable after the trial, and purchasing unlocks the full set.
Sunset Mesa Software