Setup
Set up Hebrish on Android
Eight steps, in the same order as the Setup checklist in the app. Steps 5 and 8 are optional.
The checklist
-
Install Hebrish
We recommend F-Droid: add the Hebrish repository once and new versions arrive as ordinary app updates. Or download the APK, which always gives the newest version but won't update itself.
Android 9 or newer, arm64 phones.
-
Allow the microphone
The app asks for it. If you refused twice, Android stops asking: open App info → Permissions → Microphone and choose Allow only while using the app.
-
Download a model
Jachnun (English with Hebrew vocabulary) is the default, and the one we recommend. Sabich (Hebrew plus tech vocabulary) and stock Parakeet are the alternatives; you can switch later in Models.
Each is about 660 MB, downloaded once. Use Wi-Fi.
-
Turn on the voice keyboard
Open Settings → System → Keyboard → On-screen keyboard and turn on Hebrish. The wording varies by phone; the button in the app opens the right page.
Pair it with Hebrish Keyboard or HeliBoard: their mic key opens Hebrish.
-
Floating mic Optional
A mic button over every app: tap it in a text field, talk, and the text goes in at the cursor.
On Android 13 and newer, an app installed from outside a store is "restricted" and its accessibility switch is greyed out. To unlock it:
- Open Settings → Accessibility → Downloaded apps → Hebrish floating mic and try to turn it on once. Android refuses.
- Open Hebrish's App info, tap ⋮ (top right) and choose Allow restricted settings. This item only appears after that first try.
- Go back to Accessibility and turn the floating mic on.
-
Keep it running
In App info → Battery, choose Unrestricted, so the phone doesn't stop the floating mic in the background. OnePlus phones (OxygenOS) are the most eager to stop it.
-
Turn off "Pause app activity if unused"
In App info, turn off Pause app activity if unused (Android 11 and newer; older phones call it Remove permissions if app is unused). Otherwise Android takes the microphone permission away after a few months of not opening the app, and dictation stops working.
-
All files access Optional
Only if you want the models kept in a shared folder (Models → Model location), so other on-device speech apps can use the same files. Not needed otherwise: by default the models stay in the app's own storage.
Try it
Open the app and tap Try it now to dictate into a notepad. If a word comes out wrong, tell us.
Something not working? See the FAQ or report it. Ubuntu and Windows: download page.