Troubleshooting

Start on the History page. If there is no record for the dictation, recording never started. If there is a record with a transcript but nothing was inserted, the problem is insertion. That split narrows most issues immediately.

The hotkey does nothing

• Check the profile actually has a hotkey set, and that no other profile uses the same combination. OiPer warns about both in the app.
• On macOS, Input Monitoring must be granted. See Installation.
• Another application may have claimed the combination first. Try a different one.
• If the keys need to keep working in the app underneath, enable Bypass hotkey in Settings.

Nothing is typed, but the transcript exists

• On macOS, Accessibility must be granted.
• Some applications reject synthetic paste events. Change the Insert method — direct works in places where Ctrl+V does not, and the reverse is also true.
• Enable Write to clipboard so you can paste the result manually while you narrow it down.

Recording is silent or catches nothing

• Confirm Microphone permission is granted.
• Check the Input Device settings. If you are not using the system default, the device at the top of your list may be unavailable.
• If you hear the no-speech sound, the recording contained no detectable speech — usually a muted microphone or the wrong input device.

The overlay appears in the wrong place

On macOS this needs Screen Recording permission, which is how OiPer finds the active window's position.

Transcription is slow

• Use a smaller local model. Large v3 Turbo gives close to large-model accuracy at roughly half the size; Base or Small are much faster again.
• Set Whisper Hardware to GPU if your machine has a usable one, or to CPU if GPU transcription is failing or producing bad output.
• Formatting adds a language model call to every dictation. Turning it off, or using a faster model, removes that delay.

Local transcription fails or produces nothing

The model may not be downloaded. Open the Models page and check the model your profile uses; OiPer also warns about this in the app.

Cloud dictation fails

• The model needs an API key, or an environment variable holding one. A cloud model used by a profile with no authentication is flagged in the app.
• Confirm the base URL and model name match what your provider expects.
• If a profile transcribes through a language model, that model must accept audio input.

Getting more detail

Logs are written to the logs/ folder in OiPer's application data directory:

• macOS: ~/Library/Application Support/com.oiper.desktop/logs
• Windows: %APPDATA%\com.oiper.desktop\logs

Your settings live in config.json in the same directory. Deleting that directory resets OiPer completely, including your profiles, history, and downloaded models.