Whispry, hold-to-talk voice transcription for Android
Hold-to-talk voice transcription for Android, dictated straight into any app. Volume-key hold, floating widget or keyboard button. Bring your own AI key, keep your transcripts on-device.
Android 8.0+ · sideload the apk, then Whispry updates itself
- apk downloads
- —
- latest build
- v1.1.0
- github stars
- —
- price
- free
an app built around one habittalk it out,whisprywrites it righta single trigger (volume key, floating widget or keyboard button) types a formatted transcript straight into whatever app you're in.
hold to talk
Three trigger surfaces, so dictation is never more than a press away: a volume-key hold, a draggable floating widget, or a button that rides the on-screen keyboard.
- Volume-key hold with hands-free and single/double-press modes
- Draggable floating widget that snaps to the screen edge
- Keyboard-riding trigger button, live in any text field
- Per-app suppression keeps triggers out of the way where unwanted
ai formatting
Transcription and formatting resolve independently. Groq is the default, or point either one at any OpenAI-compatible endpoint using your own API key.
- Six presets: Raw, Auto-Format, Professional, Casual, Polite, Concise
- Hinglish output: romanized Hindi instead of Devanagari
- App-aware tone: a different style per app, same dictation
- 11-language UI localization
voice commands
A spoken command router for expanding a snippet or inserting a saved value mid-dictation, plus a persistent memory the formatting step can draw on.
- Voice commands: expand, insert and app commands as the first word
- Text Expander: short typed triggers expand into full snippets
- Memory Bank & My Info: facts you never repeat yourself
- Configurable widget size, position, arming delay and idle behaviour
your key, your data
No analytics, no telemetry, and no backend of its own. Audio goes straight from your device to whichever provider you configured, using an API key only you hold.
- No analytics or telemetry, nothing about you leaves the device
- Transcripts, settings and API keys stored locally via Room + DataStore
- Audio sent device-to-provider directly, Whispry never sees it
- Open source under AGPL-3.0, in-app OTA updater from GitHub Releases
seven screens, one trigger philosophy
see it
11 languages
The interface is localized into 11 languages, not just the transcription.
bring your own key
Groq out of the box, or any OpenAI-compatible endpoint with a key only you hold.
memory bank
Persistent facts about you the formatting step can draw on, so you stop repeating yourself.
in-app updates
Whispry checks GitHub Releases itself and installs the signed APK in place.
for the people who ask what it's made of
under the hood
- ui
- Jetpack Compose with an MVI presentation layer. Each screen observes a single UiState, driven by intents.
- triggers
- The volume-key hold, floating widget and keyboard button run through an accessibility service plus overlay windows, feeding one shared capture pipeline.
- ai
- Transcription and formatting resolve independently. Groq is the default, or any OpenAI-compatible endpoint with your own key. Requests go straight from device to provider.
- data
- Room holds transcripts and memory facts, DataStore holds preferences. Hilt wires the modules together.
ready when you are
get whispry
Download the APK, allow installs from unknown sources, add your AI provider key in Settings, and you're dictating. Whispry checks GitHub for updates itself, so you never need this page again.
whispry is free and always will be
community & support
github
Source, releases and the issue tracker. Pull requests welcome.
open the reporeport a bug
Something broken or missing? File it and it gets looked at.
open an issueko-fi
Buy a coffee. Servers, test devices and late nights say thanks.
support on ko-fiupi (india)
Pay directly via UPI: cosmictaser@okicici
pay via upithe usual questions
faq
Yes. Whispry is free and open source under AGPL-3.0 (with a Section 7 attribution term). There's no subscription and no paid tier. Support is entirely optional.
Yes. Whispry ships pointed at Groq, which has a generous free tier, or you can point transcription and formatting at any OpenAI-compatible endpoint using a key you supply in Settings.
Android 8.0 (Oreo) and above.
No. Whispry is distributed as an APK through GitHub Releases. Download it from this page, allow install from unknown sources, and the built-in updater handles every version after that.
No analytics or telemetry, and there's no backend of its own. Transcripts, settings and API keys stay on-device. Audio is sent directly from your device to whichever AI provider you configured, using your own key. That provider's privacy policy governs what happens from there.
A volume-key hold (with hands-free and single/double-press modes), a draggable floating widget that snaps to the screen edge, and a button that rides the on-screen keyboard. All three are configurable and can be hidden per app.
Open an issue on GitHub, or start a Discussion. Both are linked at the bottom of this page. Pull requests are welcome too.
built, designed and maintained by
Aryan SharmaFreelance Android & software developer. Available for app builds, feature work and rescue missions.