How Superwhisper's bring your own key setup works

Superwhisper published a guide to Bring Your Own Key, which runs the AI step of dictation on a provider account you already have. Superwhisper says the feature has been in the app for a long time, so this is a walkthrough of how it works and not a new release.

Dictation happens in two steps. A voice model turns your audio into text, on your device or through a cloud provider. A language model then formats the text, sets the tone and runs your modes, locally or through your own key. You can connect OpenAI, Anthropic, Gemini, Grok or any OpenAI-compatible endpoint. With a local voice model the recording stays on your device, and only the resulting text goes to the language model if that runs in the cloud.

Enterprise accounts can also connect Azure OpenAI, Amazon Bedrock or a self-hosted endpoint. Admins add a provider, API key, model ID and URL under Custom Models in the dashboard's Configuration tab, and everyone on the team gets those models without handling a key. Turning off Allow cloud AI models limits members to local models and the ones the admin approved.