Skip to content

Choose A Transcription Provider

Open Settings → Audio to choose speech-to-text. This setting is separate from the AI provider that answers questions.

Audio settings with transcription provider, microphone, and system audio controls

ProviderWhere speech is transcribedSetup
Local (NVIDIA Parakeet)On your computerLocal model setup; no API key.
On-device (Apple Speech)On a compatible MacmacOS 26+, supported language and installed Apple language asset; no API key.
Cloud (Deepgram)DeepgramYour Deepgram API key and network access.

Allow the local model download and preparation to finish. Parakeet uses local storage and compute. Check the ready status before starting an important session.

Choose Apple Speech, select a language, and use Check or Install when the language asset needs attention. Wait until the asset is installed. Availability depends on the system and language; an unsupported language or unavailable provider needs a different selection.

The interface language in Settings → Interface does not select the Apple Speech transcription language.

Enter and validate your Deepgram API key in onboarding or Audio settings. Deepgram sends audio to its cloud service and uses your Deepgram account’s billing and limits.

Deepgram transcription setup after successful credential validation

Grant macOS permissions, choose the microphone, and test microphone speech and computer audio separately. Apply provider/device changes before a new session; restart when requested.

Local transcription keeps speech-to-text local, but cloud AI can still receive transcript text. Review data handling.