WisprTool — by FuturMinds

A free, local voice-to-text tool that transcribes your speech and types it anywhere — browser, VS Code, Notepad, Slack, or any other application. No cloud subscription required.

If you want to learn how to build tools like this yourself, subscribe to FuturMinds.

Don't want to deal with Python or code? Just download and run.

👉 Download WisprTool.exe — Windows 10/11, no installation needed.

1. Download WhisperTool.exe from the link above
2. Double-click to run
3. Hold Alt + Ctrl, speak, release — text pastes wherever your cursor is

First launch downloads the Whisper model (~104 MB) once. Everything after that is instant.

Want to see how this was built from scratch in 20 minutes? 📺 Watch the full build on YouTube

1. Install Python 3.10+ (check "Add to PATH")
2. Double-click install.bat — installs all dependencies
3. Double-click run.bat — launches WisprTool
4. Hold Alt + Ctrl (default), speak, release — text is pasted at your cursor

1. Install Python 3 or via Homebrew: brew install python3
2. Open Terminal in the project folder and run:

   chmod +x run.command
   pip3 install -r requirements.txt

3. Double-click run.command — launches WisprTool
4. Grant permissions when prompted:
   • Accessibility — for global hotkey detection System Settings → Privacy & Security → Accessibility → add Python/WisprTool
   • Microphone — for recording System Settings → Privacy & Security → Microphone → allow
5. Hold Alt + Ctrl (default), speak, release — text is pasted at your cursor

First run downloads the Whisper base model (~150 MB). Subsequent runs are instant.

If you received a WisprTool.exe file:

1. Place it in any folder
2. Double-click to run — no Python or installation needed
3. On first run, the Whisper model downloads to %USERPROFILE%\.cache\huggingface\
4. Settings are saved as config.json in the same folder as the exe

Share the single WisprTool.exe file with anyone on Windows. They need no additional software.

1. Position your cursor in any text field in any application
2. Hold the hotkey (default: Alt + Ctrl) — a circular animated overlay appears
3. Speak — the waveform responds to your voice
4. Release the hotkey — a loading spinner shows while transcribing
5. Your text is pasted at the cursor automatically

1. Click Change next to the hotkey display
2. Press and hold your desired modifier keys (e.g., Ctrl + Shift)
3. Release — the new hotkey is saved automatically

1. Tick Enable AI rewriting in the AI Rewriting section
2. Choose a provider: Gemini, OpenAI, or Claude
3. Paste your API key
4. Optionally add style/tone instructions, e.g. "Professional and concise"
5. Click Apply Settings to save (you'll see a green confirmation)
6. Every transcription is now automatically polished by the AI before pasting

If the AI call fails, the error message appears in red below the Apply button.

# 1. Install build dependencies
pip install pyinstaller

# 2. Run the build script
build.bat

The exe is created at dist\WisprTool.exe.

# 1. Install build dependencies
pip3 install pyinstaller

# 2. Make the script executable and run it
chmod +x build.sh
./build.sh

The binary is created at dist/WisprTool.

macOS app signing: Unsigned macOS binaries may be blocked by Gatekeeper. Right-click → Open the first time, or run:

xattr -dr com.apple.quarantine dist/WisprTool

wisprTool/
├── main.py              # Main application (UI, audio, transcription, AI)
├── text_processor.py    # Text post-processing (fillers, backtrack, lists, punctuation)
├── requirements.txt     # Python dependencies
│
├── install.bat          # Windows: install dependencies
├── run.bat              # Windows: launch without console window
├── run.command          # macOS: launch script
├── build.bat            # Windows: build single .exe
├── build.sh             # macOS: build single binary
│
├── VERSION              # Current version (read by the app at runtime)
├── CHANGELOG.md         # Release history
├── README.md            # This file
├── .gitignore           # Git ignore rules
│
└── config.json          # Auto-created at runtime (not committed)

Settings are saved automatically to config.json. You can also edit it manually:

{
  "hotkey": ["alt", "ctrl"],
  "model_size": "base",
  "language": "en",
  "remove_fillers": true,
  "backtrack": true,
  "numbered_lists": true,
  "smart_punctuation": true,
  "ai_rewrite": false,
  "ai_provider": "Gemini",
  "ai_api_keys": {
    "Gemini": "",
    "OpenAI": "",
    "Claude": ""
  },
  "ai_style": ""
}

Model sizes (larger = more accurate but slower): tiny · base (default) · small · medium · large-v3

Language: Set to en for English, or leave empty for auto-detect.

1. Go to aistudio.google.com/apikey
2. Create an API key and paste it in WisprTool

1. Go to platform.openai.com/api-keys
2. Create an API key and paste it in WisprTool

1. Go to console.anthropic.com
2. Create an API key and paste it in WisprTool

All AI API calls use Python's built-in urllib — no extra HTTP library needed.

WisprTool uses Semantic Versioning (MAJOR.MINOR.PATCH).

Where the version lives:

• VERSION file at the project root — single source of truth
• Shown in the app title bar: "WhisperTool v1.0.0 (by FuturMinds)"
• Bundled into the exe automatically by the build scripts

To cut a new release:

# 1. Update the VERSION file
echo "1.1.0" > VERSION

# 2. Update CHANGELOG.md with new entries under ## [1.1.0]

# 3. Commit, tag, push
git add VERSION CHANGELOG.md
git commit -m "Release v1.1.0"
git tag -a v1.1.0 -m "Release v1.1.0 — <short summary>"
git push origin master --tags

# 4. Build the exe (Windows) or binary (macOS)
build.bat        # Windows
./build.sh       # macOS

# 5. Upload dist/WisprTool.exe as a GitHub Release asset
gh release create v1.1.0 dist/WisprTool.exe --title "v1.1.0" --notes "See CHANGELOG.md"

Version bumping guide:

• PATCH (1.0.0 → 1.0.1) — bug fixes, minor tweaks
• MINOR (1.0.0 → 1.1.0) — new features, backward compatible
• MAJOR (1.0.0 → 2.0.0) — breaking changes to config or behavior

All past releases are documented in CHANGELOG.md.

No text is pasted

• Make sure your cursor is in a text field before releasing the hotkey
• On Mac, verify Accessibility permission is granted

"Permission denied" on microphone

• Mac: System Settings → Privacy → Microphone → allow Python or WisprTool
• Windows: Settings → Privacy → Microphone → allow desktop apps

Hotkey not detected

• Mac: Accessibility permission is required for pynput to detect keys globally
• Try running as administrator on Windows if using a restricted account

AI rewriting fails

• Double-check your API key
• Verify you have credits/quota on the provider's dashboard

First run is slow

• The Whisper model (~150 MB for base) downloads once on first use

Drop feature requests as GitHub Issues — the most upvoted ones get built in the next version.

If this saved you $12/month, a GitHub star takes 2 seconds and helps others find it.