yaptic

Hold a key, talk, let go. Your words appear wherever you're typing.
Push-to-talk dictation for macOS that runs Whisper on your Mac.

Why Yaptic

• Your voice stays on your Mac. Speech recognition is whisper.cpp running locally on Apple Silicon with Metal, and audio is never uploaded.
• Cleanup is optional, with the provider of your choice. It removes "um"s, fixes punctuation and resolves self-corrections ("jam 3, eh jam 4" becomes "jam 4"). Use Anthropic, OpenAI, or any OpenAI-compatible endpoint, including Ollama or LM Studio on your own machine. With cleanup off, Yaptic works fully offline.
• You can mix Indonesian and English the way you talk. Nothing gets translated.
• A custom dictionary keeps names and technical terms such as "Kubernetes" and "Tauri" spelled right.
• It works in any app. Text is pasted at the cursor, and your clipboard (images and files included) is put back afterwards.

Requirements

• A Mac with Apple Silicon (M1 or later), macOS 12 or newer.
• About 600 MB of disk space for the speech model.
• For cleanup (optional): an API key from your provider, or a local model server such as Ollama or LM Studio, which needs no key.

Install

Download

1. Get Yaptic_<version>_aarch64.dmg and SHA256SUMS.txt from Releases. Only download Yaptic from this page.

2. Check that the download is the real one. Each release is built by GitHub Actions from the tagged source and comes with checksums and a signed build-provenance attestation:

   shasum -a 256 -c SHA256SUMS.txt
   gh attestation verify Yaptic_<version>_aarch64.dmg --repo iniandrew/yaptic

3. Open the .dmg and drag Yaptic to Applications.

4. Open Yaptic. Releases aren't signed with an Apple Developer ID yet, so the first launch shows Apple could not verify "Yaptic" is free of malware… Close the message, then allow the app in one of two ways:

   • Open System Settings → Privacy & Security, scroll down to Security, click Open Anyway next to the message about Yaptic, and confirm with your password or Touch ID.

   • Or remove the download quarantine flag in Terminal:

     xattr -dr com.apple.quarantine /Applications/Yaptic.app

   You only need to do this once per version. (On macOS 14 and earlier, right-clicking the app and choosing Open also works; macOS 15 removed that shortcut.)

Build from source

You need Xcode Command Line Tools, Rust, Node.js, and CMake:

xcode-select --install
brew install rust node cmake
git clone https://github.com/iniandrew/yaptic.git
cd yaptic
npm install
npm run tauri build

The app is written to src-tauri/target/release/bundle/macos/Yaptic.app, and a .dmg sits next to it in bundle/dmg/. See CONTRIBUTING.md for running in development.

First run

1. Yaptic lives in the menu bar (the bars-and-caret icon). There is no Dock icon.
2. The settings window opens and offers to download the speech model (Whisper large-v3-turbo, about 550 MB). The download can be paused and resumed, and it is checked against a pinned SHA-256 before use.
3. Grant two permissions when macOS asks, or from the Permissions section in settings:
   • Microphone, to hear you.
   • Accessibility, to paste with ⌘V. Without it, Yaptic copies the text to your clipboard so you can paste it yourself.

Using it

Hold ⌥ Space (Option+Space), speak, and release. A small pill at the bottom of the screen shows what's happening: lime bars while listening, a wave while transcribing or cleaning up, then a check once the text is pasted. Quick taps are ignored, so you won't trigger it by accident.

Open Settings from the menu bar icon to change:

• the shortcut (any modifier plus a key, or a function key),
• the language (auto-detect, Indonesian, or English),
• your custom dictionary,
• cleanup (on/off, provider, base URL, model, API key, timeout) with a Test connection button,
• launch at login and the clipboard restore delay.

Cleanup providers

Provider	Base URL	Model	API key
Anthropic	https://api.anthropic.com	e.g. claude-haiku-5-5	required
OpenAI	https://api.openai.com/v1	a chat model name	required
Ollama	http://localhost:11434/v1	e.g. llama3.2	none
LM Studio	http://localhost:1234/v1	the loaded model's name	none

Groq, OpenRouter, Together and other OpenAI-compatible services work the same way with their base URL. Settings shows the exact endpoint each request will go to, and warns you if a remote URL uses plain http://. Local models can be slower than cloud ones, so raise the timeout if cleanup keeps falling back to the raw text.

Privacy

• Audio and the speech model never leave your Mac.
• With cleanup off, nothing is sent anywhere.
• With cleanup on, the transcript text goes to the provider and base URL you configured, under that provider's own privacy policy. Audio is never sent. With a local server such as Ollama, the text stays on your machine too.
• API keys are stored in the macOS Keychain, one per provider. They are never written to the settings file or to logs.
• After pasting, your previous clipboard is restored. The dictated text is marked as transient so clipboard managers don't keep it.

Settings live in ~/Library/Application Support/io.github.iniandrew.yaptic/.

Troubleshooting

The shortcut or paste does nothing. macOS ties permissions to a specific app build. Open System Settings → Privacy & Security → Accessibility (and Microphone), remove Yaptic, add it again, and relaunch. This is common after installing a new version. If you run from source with npm run tauri dev, the permissions belong to your terminal app instead.

The pill says "Speech model not downloaded" or the download stalls. Open Settings → Speech model and press Resume download. Partial downloads continue where they stopped.

macOS says it "could not verify Yaptic is free of malware". That's Gatekeeper blocking an app without an Apple Developer ID signature. Go to System Settings → Privacy & Security and click Open Anyway, or run xattr -dr com.apple.quarantine /Applications/Yaptic.app. See step 4 under Download.

Cleanup says it failed. The raw transcript is still pasted. Press Test connection in settings; it shows the server's real error (wrong key, unknown model, wrong URL).

Another app pastes your old clipboard. Raise the clipboard restore delay in settings, for example to 600 ms.

Limitations

• macOS on Apple Silicon only.
• The shortcut must include a key besides modifiers, so Fn alone or Right Option alone can't be used yet.
• Auto-detect costs about a second extra per dictation on an M1, because whisper detects the language in a separate pass. Picking Indonesian or English in settings is faster.
• Cloud cleanup needs internet. Offline, Yaptic pastes the raw transcript, or you can turn cleanup off or use a local model.
• Text is pasted into whichever app is in front when transcription finishes, so don't switch apps mid-dictation.
• Releases are not signed with an Apple Developer ID yet.

Contributing and security

See CONTRIBUTING.md to build and test.

License

The code is released under the MIT License. Third-party components and their licenses are listed in THIRD_PARTY_NOTICES.md.

The MIT license covers the code only. The Yaptic name and logo are not licensed for use as another product's branding. If you ship a fork, please give it a different name and icon.