koe-setup

To set up Koe for macOS voice input, you must install the application, grant microphone, Accessibility, and Input Monitoring permissions, and configure your speech recognition and text correction credentials. This ensures reliable voice-to-text pasting across applications.

Configuring Koe speech recognition requires a Volcengine ASR account for transcription and an OpenAI-compatible LLM endpoint for text correction. You need to set these as secure environment variables to enable the voice-to-text workflow.

You can generate domain-specific dictionaries for Koe voice input to improve transcription accuracy. This allows you to add technical terminology and customize correction prompts so the LLM accurately refines your speech.

Koe requires an OpenAI-compatible LLM endpoint for complete functionality because the LLM performs the language-model correction on transcribed speech. Without it, the voice input workflow cannot automatically refine and correct your text.

If Koe voice-to-text is not pasting correctly, you likely lack the required macOS Accessibility or Input Monitoring permissions. Troubleshooting involves verifying these permissions are granted and checking your hotkey and feedback settings.

The best way to change hotkeys for macOS voice input is through the Koe personalization settings. You can configure custom hotkeys and adjust feedback sounds to verify that your voice input activates reliably.