Troubleshooting on Android

The microphone records nothing#

• “Allow microphone access in the TypeWhisper app.” means the permission is missing. Open Settings → General and tap Allow microphone access. If you denied it before, the row reads Open microphone settings; turn on Microphone there (Android Settings → Apps → TypeWhisper → Permissions → Microphone).
• “The microphone delivers no sound. Is access blocked, or is another app using it?” appears while recording when the microphone sends complete silence. Check that microphone access is not blocked in Quick Settings, end any call and close other apps that use the microphone.
• If a recording ends that way, TypeWhisper shows “The microphone delivered no sound.” instead of a transcription and does not send anything to the speech engine.
• “No speech heard yet. Speak closer to the microphone.” means the microphone works but hears no voice. Speak closer to the phone. If you dictate softly, turn on Whisper mode in Settings → General.
• “No speech detected.” after you stop means the recording contained no speech. With Skip silence turned on in Settings → Speech recognition → Advanced, recordings without speech are ignored so that no text gets invented.

The speech model does not download or loads slowly#

• Model downloads run in the background and show their progress in a notification. “Queued. Starts when a connection is available.” means TypeWhisper is waiting for an internet connection.
• “Waiting to retry. Resumes automatically.” appears after an interrupted download. Wait, or tap Retry now.
• After Cancel download, the parts already downloaded are kept. Tap Resume download to continue where it stopped.
• “Download failed: …” shows the reason. Tap Retry on a stable connection and make sure the phone has enough free storage. Parakeet needs about 670 MB.
• To free storage, open Settings → Speech recognition, tap the arrow next to an engine and delete models you do not use. You can download them again later.
• “Download the speech model in TypeWhisper settings.” in the app, or “Speech model needed” in the keyboard, means the selected engine's model is not on the phone yet. Tap Download model right in the keyboard, or download it in settings.
• With Android speech recognition, “This offline language is not installed.” means Android lacks the language. Select the engine in Settings → Speech recognition and tap Download selected offline language, or choose another engine.
• By default, on-device models are freed from memory when you have not dictated for 10 minutes. This also happens when Android runs low on memory. The next dictation then takes a few seconds longer while the model loads. To change the delay, open Settings → Speech recognition → Advanced → Free model memory.

The TypeWhisper keyboard does not appear#

• Open Settings → Keyboard and tap Enable TypeWhisper Keyboard. Android opens its input settings, where you turn on TypeWhisper Keyboard.
• Then tap Switch keyboard on the same page and choose TypeWhisper as the active keyboard. On the TypeWhisper keyboard itself, the keyboard key left of the space bar switches to your next keyboard; press and hold it to pick one.
• If the keyboard shows an error that needs setup, such as a missing model or API key, tap the message to open the app.
• “Your dictation is ready. Choose where to insert it.” appears when the text field changed while the dictation was processed. Tap Insert here in the field you want, or Dismiss.
• In password and incognito fields, dictation works, but nothing is saved to History or kept for recovery. In password fields, profiles are not applied.

The dictation bubble does not show#

Open Settings → Dictation bubble and check that Show dictation bubble is on. The page then tells you what is missing, for example “Enable TypeWhisper in Android’s accessibility settings.” Tap Open accessibility settings and turn on Dictation bubble. If it says the accessibility service is disconnected, turn Dictation bubble off and on again in Android's accessibility settings. When everything works, it reads “Ready. Open a text field in another app to dictate.”

• The bubble needs microphone access. If it is missing, the same page shows Allow microphone. If you denied it before, the button reads Open microphone settings.
• The bubble only appears when you tap into an editable text field in another app. It does not show inside TypeWhisper, on the lock screen, or while a dictation from the app or the keyboard is running.
• Password fields and fields marked as sensitive are excluded. Some fields cannot accept text from the bubble: then you see “This field could not accept the dictation. Try another text field.”
• If the field changed while the dictation was processed, the text waits. Tap the paste button in a text field to insert it, or × to discard it.
• If TypeWhisper is turned off in Android's accessibility settings, the bubble disappears. Tap Allow bubble notifications on the bubble page: you then get the notification “Dictation bubble is off”, and Reactivate opens the accessibility settings. Android cannot deliver this notice if the app was force-stopped.
• Dictations in incognito fields, and in fields whose privacy mode Android cannot identify (including all fields on Android 12 and older), are inserted but not saved to History.

A cloud service or AI writing fails#

• “Set up the selected speech service in TypeWhisper settings.” means the API key is missing. Open Settings → Cloud providers, choose the provider, paste the key, tap Save and then Test.
• “The API key was rejected.” Check that you copied the whole key and that your account with the provider is active.
• “Too many requests. Try again in a moment.” The provider is limiting your requests. Wait a little and dictate again.
• “No connection to the transcription service.” Check your network. To keep dictating without a connection, choose an on-device Fallback engine in Settings → Speech recognition → Advanced.
• “Inserted without the profile: set up AI writing in TypeWhisper settings.” Open Settings → AI writing and tap Set up step by step. If the service is set up but fails, TypeWhisper inserts the plain dictation instead.
• Failed dictations are kept on the phone for 24 hours. If Settings → General → Keep recordings is not set to Delete after transcription, the last three successful ones are kept too. When there is something to recover, Recoverable recordings appears at the top of History, where you can transcribe it again, also with another engine. Dictations in password and incognito fields are not kept.

Background work stops on some phones#

Some manufacturers stop apps in the background more strictly than standard Android. On Honor phones with MagicOS, for example, TypeWhisper can only check watched folders in the background when it is allowed to start by itself.

• On Honor and Huawei phones, once you have added a watched folder, the Watched folders section on the File transcription screen shows a hint with Open App launch. Set TypeWhisper to manage manually there and keep Auto-launch and Run in background on.
• On other phones, if new files only show up when you open the app, allow TypeWhisper to start in the background in the phone's battery settings.
• Dictations you start in the app keep running in the background with an ongoing notification. Allow notifications so that you see it.