Text Cleanup & AI

Automatically polish your dictation with AI-powered grammar fixes, rephrasing, or custom prompts.

Overview

VoxInk can automatically clean up your dictation using AI. This is separate from transcription — text cleanup runs after your speech has been converted to text, polishing the output before it's pasted into your application.

Think of it as a two-step pipeline: first your voice becomes text (transcription), then that text is refined (cleanup). You can use one without the other, or both together.

Enabling Text Cleanup

  1. Click the VoxInk icon in the menu bar
  2. Go to Text Cleanup
  3. Click Enable (a checkmark appears when active)

When enabled, every dictation is processed through your chosen AI model before being pasted. When disabled, your raw transcription is pasted directly.

Cleanup Styles

VoxInk offers three cleanup styles, each providing a different level of text refinement.

Cleanup (Fix Grammar) — Default

  • Fixes punctuation and capitalisation
  • Removes filler words ("um", "uh", "like", "you know")
  • Preserves your original meaning and content
  • Light touch — minimal changes to your wording

Best for: quick dictation, messages, notes, and anywhere you want your words with just a light polish.

Rephrase (Polish)

  • Polishes grammar and sentence structure
  • Improves tone and clarity
  • More significant changes than Cleanup — may restructure sentences

Best for: professional emails, documents, reports, and anything where clarity and tone matter.

Custom Prompt

  • Write your own instructions for the AI
  • Full control over how your text is processed
  • Click Edit Custom Prompt to write or edit your instructions
  • You tell VoxInk what to do with your text, and the AI follows your instructions

Best for: specialised formatting, specific tone of voice, domain-specific requirements, or any workflow where the built-in styles don't quite fit.

Note
On the Free plan, only Cleanup and Rephrase are available. Custom prompts require a Pro plan or higher.

AI Backend Options

Choose where your text cleanup processing runs. VoxInk supports multiple backends so you can balance convenience, cost, and privacy.

Managed API (Default)

  • Uses Gemini 2.0 Flash via VoxInk's servers
  • No API key needed — works out of the box for all paid plans
  • Included with your subscription at no extra cost
  • Fast and reliable with no setup required

On This Mac (Private AI)

Run text cleanup entirely on your own computer for complete privacy. This requires a free program called LM Studio to be installed and running on your Mac.

  1. Download and install LM Studio (free)
  2. Open LM Studio and download an AI model (it will suggest one for you)
  3. Click "Start Server" in LM Studio
  4. In VoxInk, go to Text Cleanup → On This Mac
  5. VoxInk will auto-detect LM Studio. If prompted, enter: http://localhost:1234/v1/chat/completions
  6. Click Save
What is LM Studio?
LM Studio is a free program that runs AI on your Mac. It's like having ChatGPT on your computer without needing the internet. Download it once, and VoxInk can use it for text cleanup.

Cloud Providers

You can also connect VoxInk to popular AI services using your own account. Each service requires an access code (sometimes called an API key) that you create on their website.

OpenAI (ChatGPT)

  1. Go to platform.openai.com/api-keys and create an access code
  2. In VoxInk, go to Text Cleanup → OpenAI
  3. Paste your access code and click Save

Anthropic (Claude)

  1. Go to console.anthropic.com/settings/keys and create an access code
  2. In VoxInk, go to Text Cleanup → Anthropic
  3. Paste your access code and click Save

Google Gemini

  1. Go to aistudio.google.com/apikey and create an access code
  2. In VoxInk, go to Text Cleanup → Gemini
  3. Paste your access code and click Save
Note
These services charge a small amount each time they process your text. The Managed AI included with your VoxInk subscription has no extra charges.

Tips

  • Text cleanup adds a small delay (typically 0.5–2 seconds) while the AI processes your text. The exact delay depends on your backend and text length.
  • If you don't need cleanup for a quick note, you can disable it temporarily from the menu bar — no need to change your whole setup.
  • Custom prompts work best when they're clear and specific about the desired output format. Include examples if possible.
  • When using LM Studio on your Mac, start with the model it recommends. If cleanup feels slow, try a smaller model in LM Studio.
  • You can switch between backends at any time — your cleanup style preference is saved separately from your backend choice.
PreviousTranscription Setup