AiMatch

Lipwise

Lipwise turns speech into text in desktop apps and can use AI instructions to correct, format, rewrite, or translate it. It is for desktop users who want system-wide dictation with local speech recognition and a choice of local, provider-based, or no AI editing.

Visit Lipwise Compare alternatives Own this tool? Claim it
Lipwise screenshot

Who it's for

  • Writers and other desktop users who want to dictate and edit text directly in the app they are using.
  • Privacy-conscious users who want speech recognition to run locally and may also choose local AI editing or no AI editor.
  • Teams or individuals who need spoken commands to format, rewrite, or translate dictated text across Windows, macOS, or Linux apps.

Not the right fit if…

  • People who need a mobile dictation app: the listed installers are for Windows, macOS, and Linux.
  • Linux users who need compositor-independent shortcuts and typing on Wayland: the page says these depend on the compositor and that X11 works best.

How it fits your workflow

Install Lipwise, download a speech model, and choose an editor: local AI, a configured provider, or off. Hold the shortcut in the app where you want the text, speak, and release it; Lipwise transcribes locally and types the result into that app.

Use Dictate mode for speech mixed with instructions, Plain mode for verbatim dictation, or Command mode to act on selected text. Voice commands can also add structure, replace text, rewrite it, or translate it.

Audio stays on the computer. If a cloud AI provider is selected, Lipwise sends transcript text to that provider, not audio; if the provider is unavailable, the raw transcript is typed.

Pricing

Free: The app is free and MIT licensed. The page does not list paid plans.

Free

Free

  • · Open source under the MIT license
  • · Windows, macOS, and Linux installers
  • · No account required

Prices checked on Nov 9, 2026 from the vendor's site. They can change; confirm before you buy.

Key features

System-wide voice dictation
Dictate into any app using a keyboard shortcut. Lipwise types the transcription into the active app and restores the clipboard afterward.
Three dictation modes
Dictate mode accepts speech mixed with instructions, Command mode applies instructions to selected text, and Plain mode types speech without AI editing.
Local speech recognition
Choose from 69 speech models that run locally after download. The page lists model families including Parakeet, Whisper, Canary, Moonshine, SenseVoice, Qwen3-ASR, Voxtral, GigaAM, and Granite.
Local or provider-based AI editing
Use a local Qwen3.5 model through llama.cpp, configure a supported AI provider, or turn the editor off. Built-in instructions include rewriting, translation, and formatting.
Voice commands and phrasebook
Speak commands for punctuation, paragraph and list formatting, replacing text, rewriting, and translation. Users can define phrases, names, jargon, and house style.
Wake word and hands-free operation
Use a shortcut in hold-to-talk or toggle-to-talk mode, and optionally configure a wake word so only speech starting with it is treated as an instruction.

Works with

  • Claude and OpenAI
  • OpenAI-compatible endpoints
  • Ollama, LM Studio, vLLM, and llama.cpp server
  • llama.cpp and Qwen3.5

Limitations to know

  • The listed installers support Windows 10 and 11, macOS 11 or later on Apple Silicon, and Linux packages or AppImage.
  • On Linux, the page says X11 works best; shortcuts and typing on Wayland depend on the compositor.
  • On macOS, Microphone and Accessibility permissions are required for first use; Accessibility lets Lipwise type into other apps.
  • The local AI setup downloads a matching llama.cpp build and a Qwen3.5 model; listed model sizes are 2.7 GB for 4B and 5.7 GB for 9B.

Alternatives to consider

  • Spokenly

    Choose it if you also need dictation on iPhone, which Spokenly supports in addition to desktop platforms.

  • VoiceInk

    Choose it if you specifically need a macOS dictation app with personal vocabulary and text replacements.

Compare Lipwise side by side

Getting started

Windows 10 or 11; macOS 11 or later on Apple Silicon; or Linux. macOS requires Microphone and Accessibility permissions on first use. Linux X11 works best; shortcuts and typing on Wayland depend on the compositor.

  1. Download the installer for your platform from GitHub Releases.
  2. Install Lipwise and grant the required macOS permissions if prompted.
  3. Choose a speech model and an editor in the app.
  4. Hold Ctrl Shift Space in any app, speak, and let go.

Frequently asked questions

Does Lipwise work offline?

Yes. Once a speech model is downloaded, transcription works fully offline. With the local editor or no editor, the rest of the workflow can also work offline.

What if Lipwise mistakes my words for an instruction?

Use Plain mode for verbatim dictation, or configure a wake word so only speech starting with it is treated as an instruction. History shows what was heard alongside what was typed.

Which editor should I choose?

The page recommends On This Computer for privacy and says the 4B model handles built-in instructions. For long rewrites and translations, it recommends Claude or the 9B local model.

How does Lipwise update?

It checks GitHub Releases, downloads updates quietly, and installs them on the next restart. Update checks can be disabled in Settings.

Can I build Lipwise myself?

Yes. The page describes a Tauri 2 app with a Rust core and React interface, and says CONTRIBUTING.md covers building, tests, and the project layout.