Set up your API keys (BYOK)
Local dictation needs no key. VoiceWink works out of the box with local models — your chosen model downloads once during setup; no API key, no account, nothing to sign up for. Keys only matter if you opt into the cloud extras: dictation on the big cloud models, AI enhancement, and voice-to-image.
With bring-your-own-key there is no middleman: requests go straight from your PC to the provider on your own account, and you pay the provider's own (usually tiny, pay-as-you-go) prices. Your key is pasted once and encrypted on your machine by Windows (DPAPI) — it is never sent to DV Studio.
Four steps
- 1. Pick a provider from the table below and create a free account.
- 2. Create an API key in the provider's console (the links go straight there).
- 3. Paste it into VoiceWink — transcription keys on the Models page, AI enhancement and image keys on the AI Enhancement page.
- 4. Dictate. Pick the cloud model in the same place; you can switch back to local any time.
Providers
| Provider | Used for | Get your key |
|---|---|---|
| OpenAI | Transcription · AI enhancement · Images | platform.openai.com/api-keys |
| Groq | Transcription · AI enhancement | console.groq.com/keys |
| Deepgram | Transcription | console.deepgram.com |
| ElevenLabs | Transcription | elevenlabs.io → API keys |
| Anthropic | AI enhancement | console.anthropic.com → keys |
| Google Gemini | AI enhancement · Images | aistudio.google.com/apikey |
| Mistral | AI enhancement | console.mistral.ai/api-keys |
| OpenRouter | AI enhancement · Images | openrouter.ai/settings/keys |
| Cerebras | AI enhancement | cloud.cerebras.ai |
Running your own models? For AI enhancement, any OpenAI-compatible endpoint works — point the custom base URL at it (local Ollama included). With a local endpoint, your prompts never leave your machine.
Good to know
- Keys are stored encrypted (Windows DPAPI) under your user profile and are excluded from settings exports and problem reports.
- Usage is billed by the provider to you directly — VoiceWink adds no markup and has no quota of its own.
- You can mix and match: local transcription with cloud enhancement, or the other way round.
Stuck? support@dvstudio.app