Your models, your keys, your network.
Today we’re telling you about a Superwhisper feature that has been there for a long time: Bring Your Own Key (BYOK). Individuals and organizations can run Superwhisper’s AI step on the provider they already trust. Pair it with a local voice model, and your audio never leaves your device.
You pick the provider, the model and where it runs, and your existing agreements decide how your data is handled.
What is BYOK for dictation?
BYOK connects Superwhisper to your own AI provider account through an API key, so the language model step of your dictation runs on your account. Connect OpenAI, Anthropic (Claude), Gemini, Grok or any OpenAI-compatible endpoint.
On Enterprise, you can also connect Azure OpenAI, Amazon Bedrock or a self-hosted endpoint to keep processing in the region you choose.
See it in action
Ryan shows how to add your own API keys and how to connect a language model you host yourself.
Local transcription. Custom LLM post-processing. Your own keys.
How does BYOK work in Superwhisper?
Superwhisper handles dictation in two steps, and each step can run in a different place.
- Voice model: turns your audio into text, on your device or through a cloud provider.
- Language model: formats your text, sets the tone and runs your modes, locally or through your own key (BYOK).
Pick a local voice model and the recording never leaves your device, even when your language model runs in the cloud.
Why enterprise teams use BYOK for dictation
Most companies already have approved AI providers and signed data terms, and BYOK lets Superwhisper run on those accounts. Admins add the approved models once, and everyone on the team gets them without managing a key.
Admins can also turn off Superwhisper’s hosted cloud models, which limits members to local models and the custom models you approved.
Does audio go to the cloud?
It depends on the voice model you choose. With a local voice model, the recording stays on your device, and only the resulting text goes to your language model if it runs in the cloud.
With a cloud voice model, audio goes to the provider that does the transcription, under your agreement with that provider. Your transcription history stays on your device.
What makes Superwhisper BYOK different?
Admins push shared credentials to the whole team, limit members to approved models and switch off the hosted cloud. Your team dictates in any app, and IT decides which models process the words.
Get started
Setup takes three steps in your dashboard. Open the Configuration tab, add your provider, API key, model ID and URL under Custom Models, then turn off Allow cloud AI models if members should use only your models.
For yourself, pick a Superwhisper plan and add your key in settings. For your organization, get started with Superwhisper for teams.