Configuring AI Enhancements
Connect to AI providers and configure enhancement behavior.
What is AI Enhancement?
AI Enhancement takes your raw transcription and sends it to a Large Language Model (LLM) for post-processing. This lets you automatically reformat text, fix grammar, summarize content, or change the tone of your writing. You can create different prompts for different tasks — a professional email prompt, a casual message prompt, a code comment prompt, and so on.
To use this feature, enable it and configure an AI provider.
Enabling Enhancement
- Navigate to the Enhancement tab.
- Toggle on the Enable Enhancement switch at the top.
You can also toggle enhancement on and off quickly with ⌘E while the Mini Recorder is visible. See Shortcuts for all keyboard shortcuts.
Connecting to an AI Provider
Supported Providers
- Ollama (Free & Local) — Run open-source models locally. Best for privacy and offline use.
- Groq (Free Tier) — Very fast inference speeds.
- Cerebras (Free Tier) — High-speed inference with generous free limits.
- OpenAI (Paid) — GPT-4o and other OpenAI models.
- Gemini (Free Tier) — Google's Gemini models, with a generous free tier.
- Anthropic (Paid) — Claude family of models.
- Mistral (Paid) — Mistral AI models.
- OpenRouter (Paid) — Access many providers through a single API key.
- Custom — Connect to any OpenAI-compatible API endpoint.
How to Configure a Provider
- In the Enhancement tab, find the AI Provider Integration section.
- Select your provider from the dropdown.
- Enter your API key. A green "Connected" status confirms the key is valid.
- Select a model from the list that appears once connected.
Using Ollama for Local Enhancement
- Install and run the Ollama application on your Mac.
- Select Ollama as your provider in VoiceInk.
- VoiceInk connects automatically to
http://localhost:11434. - Select any model you have pulled in Ollama.
Enhancement Settings
Click the gear icon in the Enhancement tab to access additional settings.
Clipboard Context
When enabled, VoiceInk reads the current contents of your clipboard and passes it to the AI as additional context. This is useful when you want the AI to rewrite or respond to something you've copied.
Screen Context
When enabled, VoiceInk takes a one-time screenshot of your active window, extracts the visible text using on-device OCR, and passes it to the AI as context. Your screen is not recorded continuously, and the image itself is never sent to any provider — only the extracted text.
For full details on how this works and privacy implications, see Contextual Awareness.
Skip Short Transcriptions
When enabled, very short transcriptions are sent to the AI as-is without running enhancement. You can set the Minimum words threshold (1–15 words, default is 3). Transcriptions shorter than this threshold skip enhancement entirely.
This avoids wasting API calls on single-word dictations or accidental recordings.
Enhancement Shortcuts
- ⌘E — Toggle AI Enhancement on/off while the Mini Recorder is visible
- ⌘1 – ⌘0 — Switch to enhancement prompt 1–10
See Shortcuts for the full shortcuts reference.
Enhancement Prompts
Your prompts are shown as a grid in the Enhancement tab. You can:
- Add a prompt with the + button
- Reorder by dragging prompts
- Edit by double-clicking a prompt
- Right-click for options (duplicate, delete, set as default)
Quickly Switching Prompts
Press ⌘1–⌘0 while the Mini Recorder is open to switch prompts instantly without opening the app. See Switching Enhancement Prompts for details.
Trigger Words
You can assign keywords to any prompt so it activates automatically when that word appears at the start or end of your speech — no keyboard shortcut needed. See Trigger Words.
Assistant Mode
Enhancement prompts can be set to Assistant Mode, which treats your speech as a question or command directed at the AI rather than text to reformat. See Assistant Mode.