Requirements
- Apple-silicon Mac with macOS 15 or later.
- Microphone permission when microphone input is selected.
- System Audio Recording permission when system audio is selected.
- Internet access for GPT Realtime, custom remote endpoints and initial local-model downloads.
- Additional free storage and unified memory for optional Qwen models.
Audio permission problems
Open System Settings → Privacy & Security, then confirm the permission required by the selected input. Restart Übersetz after changing permission if macOS does not immediately expose the new state.
System Audio Recording permission lets the app receive audio through a Core Audio tap. Übersetz does not create a screen-capture stream and does not receive screen pixels.
GPT Realtime does not start
Confirm that your own OpenAI API key is present in Settings, that the OpenAI account can use the selected API and that billing or usage limits have not blocked the request. The key is stored in macOS Keychain; never include it in a support message.
A local model is unavailable
Check free disk space and the model's memory guidance, then retry the model download. Übersetz verifies downloaded model size and SHA-256 before installation. Larger Qwen presets require substantially more unified memory and storage than the smaller presets.
Contact support
Email [email protected]. Include the app version, macOS version, Mac model, input source, translation backend and steps that reproduce the issue. Do not attach private transcripts unless you deliberately choose to share them, and remove confidential content first.