Set Up On-Device Speech Recognition
Configure Maise as your Android speech recognizer and enable voice dictation in any text field. Everything runs on-device using distil-Whisper — no microphone data is sent to any server.
Overview
Maise provides two ways to use on-device speech recognition on Android:
- System speech recognizer — Registers Maise as Android's preferred RecognitionService. Apps that call the standard SpeechRecognizer API use Maise automatically.
- Voice keyboard (IME) — Adds a voice dictation button to the keyboard in any text field across the entire system.
Both are optional and independent — you can enable either or both. The ASR model used is distil-whisper/distil-small.en, a compact English-language Whisper variant that loads quickly and runs efficiently on mobile hardware.
Set Maise as the default speech recognizer
Enable the voice keyboard
The Maise keyboard input method (IME) adds a microphone button to any text field on the device, allowing you to dictate text anywhere without switching apps.
Test it in the app
Before relying on ASR system-wide, test it in the Maise app to confirm it works on your device:
Troubleshooting
Status stays on "Loading model…" for a long time
- The Whisper model loads from storage on first use. On slower devices this can take 5–10 seconds. Subsequent loads are faster.
Shows "(no speech detected)"
- Make sure the microphone permission is granted to Maise in Android Settings → Apps → Maise → Permissions.
- Speak clearly and within range of the microphone. Background noise can interfere with VAD.
Maise does not appear as a keyboard option
- Make sure you enabled Maise in Manage keyboards (step 3 above).
- The keyboard switcher icon only appears in text fields — look for it when your keyboard is open.