Flick

Speak, rewrite, and reply anywhere.

A small desktop assistant that works in the apps you already use.

Website · Downloads · Setup guide · Privacy

Flick helps you dictate, improve text, and draft replies without moving your work into another editor. Choose private local speech models, or connect your own cloud transcription provider when you want one.

Flick 2.0.1 is a release candidate. The code and unsigned build candidates are available now; public release still requires signed and notarized installers plus final testing on physical devices. See the release checklist.

What can I do with Flick?

I want to…	What to do
Speak instead of type	Press your dictation shortcut, wait for Recording, speak, then press it again.
Improve a sentence	Add a command such as !fix or !shorter after the text.
Draft a reply	Select a message, press your reply shortcut, choose a tone, and review the draft.

Flick stays out of the way until you use a shortcut or command. The dictation and transformation pill can sit at the top or bottom of your display, and the reply assistant opens as a compact companion near your cursor.

Quick start

1. Install the right build

Open Downloads and match the file to your computer:

• Windows: choose x64 for most Intel/AMD PCs or arm64 for Snapdragon/Windows on ARM devices.
• macOS: choose Apple Silicon for M-series Macs or Intel for older Macs.
• Linux: choose the x64 package for your distribution.

Approve microphone and accessibility/input permissions when your operating system asks.

2. Choose how dictation runs

Open Settings → Models and download a small local model. Whisper Tiny English is a good first test. Flick shows the language support and download size before you choose it.

If you prefer cloud transcription, open Settings → Dictate, choose the cloud option, and enter your own OpenAI-compatible endpoint, model, and API key. Flick never falls back from local to cloud by itself.

3. Set your shortcuts

Open Settings → Write to record a reply shortcut and Settings → Dictate to record a dictation shortcut. You can also customize Copy Last Result and Paste as Plain Text under Advanced. Flick rejects shortcuts that conflict with another Flick action.

4. Try your first dictation

1. Click a normal text field in any supported app.
2. Press the dictation shortcut.
3. Wait until the pill says Recording, then speak.
4. Press the shortcut again, or release it in push-to-talk mode.
5. Flick transcribes and returns the text to the field you started from.

Press Esc while recording to discard the audio. Change the pill position under Settings → Dictate → Floating pill position.

Improve text

Write normally, then add a command at the end:

Command	Result
!fix	Fix spelling, grammar, and punctuation.
!formal	Use a more professional tone.
!casual	Make the writing friendlier.
!shorter	Make the text more concise.
!longer	Add useful detail.
!rephrase	Say the same thing more clearly.
!bullet	Turn the text into a list.
!translate:spanish	Translate to the language you name.

Create your own commands under Settings → Commands.

Draft a reply

1. Select only the message or text you want Flick to use as context.
2. Press your reply shortcut.
3. Choose a tone and describe what you want to say.
4. Generate the reply and edit it if needed.
5. Choose Copy or Insert into app.

Flick does not scan your screen or read whole conversations. It uses only the text you deliberately select, and it checks that you are still in the original app before inserting anything.

Dictate

Local models

Flick’s catalog contains 370 pinned model downloads from Hugging Face, including Whisper, Parakeet TDT/CTC/RNNT, Moonshine, Canary, Qwen3 ASR, SenseVoice, and other families supported by the speech engine. Search by family or show every available quantization when you want more control.

Each listed file has a pinned source revision, expected size, and SHA-256 hash. Downloads use a temporary partial file, resume only when the server confirms the requested range, and become selectable only after verification. Every model page links to its source and license.

The catalog is broader than the release acceptance set: not every model has been tested for speed and accuracy on every device. Release builds run real audio through representative Whisper, Moonshine, and Parakeet models on supported targets; Flick turns engine failures into recoverable errors instead of allowing the desktop app to crash.

Cloud models

Cloud dictation is optional and OpenAI-compatible. Flick sends audio only when you choose the cloud provider and start a dictation. The transcription key is separate from your writing-provider key and is stored in your operating system keychain.

Activation modes

• Toggle: press once to start and once to stop.
• Push to talk: hold the shortcut while speaking.
• Hold or toggle: tap for toggle behavior or hold for push to talk.

Shortcuts

These are the defaults. You can replace them in Settings.

Action	Windows / Linux	macOS
Reply assistant	Ctrl+Shift+Space	Cmd+Shift+Space
Dictation	Ctrl+Space	Cmd+Space
Copy last result	Ctrl+Alt+C	Cmd+Alt+C
Paste as plain text	Ctrl+Alt+V	Cmd+Alt+V

If something is not working

• The microphone does not start: allow microphone access, then check the selected device under Dictate.
• A shortcut does nothing: record it again in Settings and make sure another Flick action does not use it.
• Nothing was pasted: Flick refuses to paste after you switch apps or into a protected field. Use Copy last result instead.
• A model download stopped: choose Download again. Flick will safely resume or restart it and verify the whole file.
• A model cannot be used: remove the failed download and try the recommended smaller model. The app remains running even if the native engine fails.
• Cloud generation fails: check the provider endpoint, model name, and key under Write or Dictate.
• Wayland blocks the shortcut: bind your desktop shortcut to one of the Flick commands listed below.

If the issue continues, export redacted diagnostics from Settings and attach them to a GitHub issue. Diagnostics exclude API keys, clipboard text, prompts, replies, and transcript contents.

Privacy

• Local dictation keeps audio on your device.
• Cloud dictation is opt-in and uses only the provider you configure.
• Reply context comes only from your explicit text selection.
• API keys are stored in the operating system keychain.
• Flick refuses protected password fields and apps you exclude.
• Recordings and history are off by default and can be deleted from Settings.
• Flick has no telemetry by default.

Optional AI cleanup sends only the finished transcript—not the audio—to your configured writing provider.

Desktop commands

Desktop shortcut tools can send a fixed action to a running Flick instance:

flick --open-settings
flick --open-composer
flick --toggle-dictation
flick --cancel-dictation
flick --copy-last-result

The command line does not accept text, prompts, or shell commands.

Linux notes

The recording overlay is hidden by default on Linux because some compositors can take focus from the active text field. Flick prefers xdotool on X11 and wtype or dotool on Wayland when available. Restricted Wayland desktops may require a desktop shortcut that calls the Flick command line.

Development and verification

Install Node.js 20+, Rust 1.77+, and the Tauri v2 prerequisites.

git clone https://github.com/rixabhh/flick.git
cd flick
npm ci
npm run tauri dev

npm test
npm run test:e2e
npm run build
cargo test --manifest-path src-tauri/Cargo.toml
cargo clippy --manifest-path src-tauri/Cargo.toml -- -D warnings

CI has three clear responsibilities: Verify checks source and tests, Release builds and smoke-tests installers without publishing by default, and GitHub’s managed workflow deploys Pages. A release draft is created only after every platform succeeds.

The repository structure is straightforward:

• src-tauri/src/ — Rust desktop services, speech, models, shortcuts, and native integration.
• src/lib/ — Svelte settings, reply assistant, overlays, and UI helpers.
• docs/ — the GitHub Pages product site.
• .github/workflows/ — Verify and Release automation.
• RELEASE_CHECKLIST.md — signing, notarization, hardware, and promotion gates.
• MARKET_READY_PLAN.md — the product hardening plan and acceptance criteria.

License

Flick is available under the MIT License. Model files keep their own upstream licenses; the Models screen links to each one.