WonderWhisper is an Android dictation app that brings a floating voice-to-text button to any text field. It combines cloud transcription providers, AI cleanup, command-mode editing, custom vocabulary, and accessibility-based text insertion so you can dictate into messaging apps, notes, email, browsers, and other Android apps.
This project is now open source under the MIT License. Bring your own API keys for cloud services.
- System-wide floating dictation bubble
- Accessibility-based text field detection and insertion
- Simple Mode with recommended defaults
- Pro Mode with model/provider selection
- Streaming and non-streaming transcription flows
- AI cleanup and post-processing after transcription
- Command Mode for editing, rewriting, answering questions, and transforming selected or copied text
- Custom vocabulary replacements
- Secure local API key storage with AndroidX Security Crypto
- Activity logs, audio file management, and debugging tools
- Optional Cloudflare Worker Groq proxy scaffold for private/self-hosted builds
Transcription providers include:
- Groq Whisper
- OpenAI Whisper / GPT-4o transcription
- Google Gemini
- ElevenLabs Scribe
- AssemblyAI
- Deepgram
- Mistral Voxtral
- Soniox streaming transcription
AI cleanup providers include:
- Groq
- OpenAI
- Anthropic Claude
- Mistral
- OpenRouter
Provider availability depends on the app settings, API keys, and current provider APIs.
WonderWhisper does not ship with public provider API keys. Cloud transcription and AI cleanup require user-provided API keys unless you build your own hosted proxy.
Add keys inside the app from API Keys. Useful starting points:
Do not commit API keys, keystores, release artifacts, .dev.vars, local.properties, or Gradle properties containing secrets.
- Enable the required Android permissions: microphone, overlay, notification, and accessibility service.
- Tap the floating bubble to start recording.
- Tap again to stop recording.
- The selected transcription provider converts speech to text.
- Optional AI cleanup rewrites or formats the text.
- The accessibility service inserts the result into the active text field.
Command Mode activates when dictation starts with "command". It can rewrite selected text, use clipboard context, answer questions, or perform text transformations.
Requirements:
- Android Studio
- JDK compatible with the Android Gradle Plugin
- Android SDK 35
- NDK 28.2.13676358
Build commands:
./gradlew assembleDebug
./gradlew testDebugUnitTest
./gradlew lintInstall on a connected device or emulator:
./gradlew installDebugLocal proxy configuration is optional and should stay local:
groq.proxy.baseUrl=
groq.proxy.appToken=These values can live in local.properties or private Gradle properties. They are intentionally ignored by git.
app/- Android app moduleapp/src/main/java/com/slumdog88/dictationkeyboardai/- app sourceapp/src/main/java/com/slumdog88/dictationkeyboardai/ui/screens/- Compose screensapp/src/main/java/com/slumdog88/dictationkeyboardai/transcription/- transcription flowsapp/src/main/java/com/slumdog88/dictationkeyboardai/transcription/streaming/- streaming dictation scaffoldingworkers/groq-proxy/- optional Cloudflare Worker proxy scaffold
WonderWhisper is designed around local control:
- API keys are stored locally using encrypted storage.
- Local logs and audio files remain on the device unless you export or share them.
- Cloud providers receive audio/text only when you choose a cloud-backed transcription or AI model.
- Accessibility permissions are used to detect editable fields and insert text.
Always review provider terms and privacy policies before using third-party AI services.
Contributions are welcome. Please read CONTRIBUTING.md before opening a pull request.
Good first areas:
- Provider bug fixes
- Device compatibility fixes
- Accessibility reliability improvements
- Documentation
- Tests around settings, provider selection, and command-mode behavior
Please do not open public issues for sensitive security reports or leaked keys. See SECURITY.md.
WonderWhisper is released under the MIT License.

