Whistle turns speech into text on your phone — no account, no cloud, no upload.
Tap Start recording, speak, then tap Stop. Your words appear as editable text you can copy or share. Everything runs on-device with a small 17 MB speech model, so it works on a plane, in a meeting, or anywhere you would rather not send your voice to a server.
WHY WHISTLE
• Private by design — audio is transcribed on your phone and never uploaded
• 7 languages — English, German, French, Spanish, Italian, Dutch and Polish, auto-detected
• Offline after setup — one 17 MB model download from huggingface.co, then dictate without a connection
• Fast — 17 MB model, live or batch transcription
• No account, no subscription, no ads, no tracking
• Optional Whisper Base — 148 MB, ~99 languages, most precise; downloaded only when you choose it
• Optional system dictation — a zero-download Android on-device engine
• History stays on your phone — copy, share or delete individual notes, including kept audio
• Low-confidence words are underlined; tap one to choose a spelling suggestion or type the correction
• Custom words for names and jargon
HOW IT WORKS
1. Install the app; the 17 MB speech model downloads once from huggingface.co (Wi-Fi recommended).
2. Allow microphone access when asked.
3. Tap Start recording, speak, then tap Stop. Recording is active only between those taps.
WHO IT IS FOR
Notes and reminders, message and email drafting, meeting notes, captions and anywhere you want fast, private voice typing.
Whistle is built with the open-source Cactus Whistle model and engine (Apache-2.0).
