Rewrite selected text anywhere on Windows, privately and locally.
Improve writing, adjust its length, summarize it, or change its tone with a local small language model.
No account, API key, or cloud text processing is required.
TextRecast is a lightweight Windows desktop utility that rewrites text inside the applications you already use. Select editable text, click the floating TextRecast button, choose an operation, review or edit the local result, and replace the original selection only when it is ready.
Inference runs locally through LLamaSharp's CPU backend. Guided setup presents four cataloged models and downloads only the compatible option the user explicitly selects. No model is bundled with the installer.
- Improve grammar, spelling, punctuation, wording, and clarity
- Make text shorter or longer while preserving its meaning
- Summarize long selections
- Change tone to Professional, Casual, Friendly, Formal, or Direct
- Work in browsers, editors, messaging apps, and other Windows applications
- Run formatting locally without an account or API key
- Review and edit generated text before replacing the source selection
- Copy or regenerate a revision without changing the source text
- Choose a compatible local model before any download begins
- Resume an interrupted model download instead of starting over
- Cancel an active generation while preserving the captured source
- Revalidate the source window and selected text before replacing anything
- Launch the floating button at Windows sign-in when that setup option is enabled
- Windows 11 on x64 hardware, or a Windows 10 edition still supported by .NET 10
- The current Microsoft Visual C++ v14 Redistributable for x64
- Approximately 2-4 GB of available disk space, depending on the selected model
- An internet connection when no valid packaged or previously installed model is available
The Windows installer and installed application are self-contained. Users do not need to install .NET 10.
- Download
TextRecast-vX.Y.Z-win-x64-setup.exeand its matching.exe.sha256file from the Releases page. - Optionally verify the checksum with
Get-FileHash .\TextRecast-vX.Y.Z-win-x64-setup.exe -Algorithm SHA256. - Run Setup, review the Apache-2.0 license and privacy notice, choose the per-user destination, and select one compatible model.
- Choose Install. Setup installs the application, downloads only the selected model, and verifies its exact size and SHA-256.
- On the completion page, keep or clear the default-checked Launch, Desktop shortcut, and Start at sign-in choices, then choose Finish.
Installation does not request administrator privileges. Application files are installed for the current Windows user; downloaded models and setup state remain under that user's local application data.
TextRecast shows the available models before making a network request. Setup presents the model name, profile, and download size, disables choices that do not meet measured hardware requirements, and shows the selected model again before installation. Only that model is downloaded.
| Role | Model | Download | Status |
|---|---|---|---|
| Fast/default | Qwen 2.5 1.5B Q4_K_M | 1.04 GiB | Established default; review all output |
| Balanced | Qwen 3.5 2B Q5_K_M | 1.34 GiB | Experimental; summaries need extra review |
| Best quality | Qwen 3.5 4B Q5_K_M | 2.93 GiB | Experimental; verify roles, titles, and deadlines |
| Alternative | Granite 4.1 3B Q5_K_M | 2.27 GiB | Experimental; higher semantic-drift risk |
These profiles are currently supported for English rewriting only. None is perfect, so review every generated revision before replacement. Hardware checks can disable an unsafe choice, but they never preselect a model; Qwen 2.5 remains selectable when hardware inspection is unavailable.
Cancelling or losing the connection keeps a validated partial download. Select Retry on the same computer to continue from the saved point. TextRecast restarts safely if the server no longer accepts the saved range or the remote file has changed. A model becomes active only after its exact size and SHA-256 checksum pass verification. A valid installed selection is reused on later launches.
The first formatting request loads the model into memory. Later requests reuse the loaded model and usually start faster.
Where is the downloaded model stored?
TextRecast stores the model for the current Windows user under:
%LOCALAPPDATA%\TextRecast\Models
%LOCALAPPDATA% is a standard Windows environment variable. It expands automatically to the signed-in user's local application-data folder; it is not a path tied to the developer's computer.
For exact filenames, checksums, pinned upstream revisions, and offline setup instructions, see Models/README.md.
Why does this README contain file paths?
Only portable paths are used:
- Paths such as
src/TextRecast.App/...are relative to the repository and work wherever the project is cloned. The relative icon path at the top lets GitHub display the TextRecast icon. %LOCALAPPDATA%\TextRecast\Modelsis the Windows per-user storage location for the downloaded model.- Project paths in developer commands identify which project .NET should build or run.
There are no author-specific absolute paths in this README.
- Select editable text in another Windows application.
- Click the floating TextRecast icon.
- Choose Improve, Shorter, Longer, Summarize, or Tone.
- Choose a style when using Tone.
- Review or edit the generated revision.
- Choose Copy, Regenerate, Cancel, or Replace selection.
Replace selection returns focus to the source application, confirms that the original selection is still valid, and replaces it with the exact revised text. A clean success closes the editor. A failure, cancellation, or clipboard-restoration warning keeps the editor and revision visible with a message.
Clicking the floating icon again closes an open editor without capturing text again. Drag the icon to reposition it. Right-click it to format a new selection, view model/legal/about information, or exit TextRecast.
The complete data-handling statement is available in the TextRecast Privacy Notice.
- TextRecast itself does not write selected or generated text to logs or files.
- Formatting runs on the local CPU and does not use a text-processing API.
- TextRecast does not maintain a formatting history.
- Selections are limited to 32,000 characters and expire after 15 minutes.
- The source window, process, capture age, and selected text are checked before replacement.
- Empty, oversized, or unsafe generated output is rejected.
- Clipboard restoration is best-effort because other applications can temporarily lock the Windows clipboard.
Capture fallback and replacement temporarily use the Windows clipboard. Windows clipboard history, clipboard sync, or third-party clipboard managers may retain that content, so disable those features when working with sensitive text.
Network access begins only after the user confirms a model that is not already packaged or installed. After that download, normal formatting does not require a cloud service.
Select text in another application
|
v
Capture the selection
(UI Automation or clipboard fallback)
|
v
Format with the local SLM
|
v
Review or edit the revision
|
v
Revalidate the original selection
|
v
Replace the text and restore the clipboard when possible
The cloned project says that .NET 10 is missing
A source checkout requires the .NET 10 SDK for development. The published self-contained installer does not require a separate .NET installation.
The application reports a missing Visual C++ DLL
Install the latest x64 package from Microsoft's supported Visual C++ Redistributable downloads, then restart TextRecast.
The model download or integrity check failed
Check the internet connection and available disk space, then retry. TextRecast keeps a safe partial download after cancellation or a temporary network failure and resumes it automatically. Invalid partial data is discarded instead of being installed. Offline setup details and expected checksums are in Models/README.md.
No selected text was captured
Keep the source application open and the text selected before clicking TextRecast. The application must support UI Automation text selection or copying with Ctrl+C.
The selection could not be replaced
The window, process, selection, or capture age may have changed. The editor keeps the generated revision visible; select the source text again and start a new capture. Windows can also block simulated input when the source application is running as administrator and TextRecast is not.
Formatting is slow
Inference is CPU-only. The first request also verifies and loads the model, and large selections can require multiple inference passes. An active operation can be cancelled from the result window.
Building requires the .NET 10 SDK. From the repository root:
dotnet restore TextRecast.slnx
dotnet build TextRecast.slnx -c Release --no-restore
dotnet test TextRecast.slnx -c Release --no-build
dotnet run --project src/TextRecast.App/TextRecast.App.csprojdotnet publish src/TextRecast.App/TextRecast.App.csproj -c Release -r win-x64 --self-contained true --no-restore -o artifacts/publish/TextRecast-win-x64
./tools/Build-WindowsInstaller.ps1 -Version 0.3.0 -NoRestoreThe self-contained application publish directory is an internal installer input, not a separately supported portable release. Normal publishes intentionally contain no model; the user chooses one during setup. An offline build can explicitly include an exact cataloged GGUF file with -p:IncludeBundledModels=true; see Models/README.md.
For tagged releases, GitHub Actions creates the model-free setup executable and matching SHA-256 file. The installer contains the complete runtime and required TextRecast, .NET, WPF, and third-party legal notices. A release tag must use vMAJOR.MINOR.PATCH and match the version in Directory.Build.props.
- TextRecast.App contains the WPF presentation layer and application composition.
- TextRecast.Core contains platform-independent workflows, contracts, results, and models.
- TextRecast.Deployment contains the model catalog, verified downloads, hardware compatibility checks, setup state, and installer identity without inference dependencies.
- TextRecast.Infrastructure contains local SLM inference and Windows capture/replacement integrations.
- TextRecast.Setup contains the self-contained managed Windows setup host.
- installer contains the internal per-user application package definition.
- tests contains Core, Deployment, Infrastructure, Setup, and application UI automated tests.
- Models documents supported models, integrity metadata, and offline packaging.
TextRecast.Setup -------> TextRecast.Deployment
TextRecast.App ---------> TextRecast.Core + TextRecast.Deployment + TextRecast.Infrastructure
TextRecast.Infrastructure -> TextRecast.Core + TextRecast.Deployment
Model profiles are defined in SlmModelCatalog.cs. A new model can reuse installation, verification, storage, inference lifecycle, chunking, and workflow components; models with a different chat format should provide a matching prompt builder. The model-free setup build and its transaction boundaries are documented in installer/README.md.
- WPF on .NET 10
- NSIS 3.12 for the internal per-user application package
- LLamaSharp with CPU inference
- Qwen 2.5 GGUF, Qwen 3.5 GGUF, and Granite 4.1 GGUF
- Windows UI Automation, clipboard, and input APIs
Report security issues using the process in SECURITY.md. Data handling is described in PRIVACY.md.
TextRecast source code and documentation are licensed under the Apache License 2.0. Copyright 2026 snss10.
The TextRecast name, icon, and other brand assets identify this project. The Apache License 2.0 does not grant trademark rights beyond the uses permitted by Section 6 of the license.
See NOTICE for project attribution and THIRD-PARTY-NOTICES.md for the separate licenses and notices that apply to dependencies, icon artwork, and the downloaded model.