Which VoiceInk model to choose, from VoiceInk's own guidance
By Ryleigh Newman · Published
VoiceInk's docs name Parakeet V3 as the starting point for local dictation on Apple silicon, and it is the model onboarding sets up. For English only, they point to Parakeet V2 or Parakeet Unified. For other languages, compare Whisper, Cohere Transcribe, Nemotron and SenseVoice against the languages each supports. On an Intel Mac, the app warns that local models are not reliable and suggests a cloud provider.
I built YapToText, a different dictation app for the Mac, so read what I say about VoiceInk with that in mind. Every fact about VoiceInk here comes from its own documentation, read on October 9, 2026. VoiceInk's docs do not rank models by accuracy, and neither do I.
This app is completely free. You don't have to make an account. There are no subscriptions, no difficult downloads, and nothing is locked. Made for Apple silicon Macs. Source on GitHub.
This guide in other languages: العربية, Čeština, Dansk, Deutsch, Ελληνικά, Español, Suomi, Français, עברית, हिन्दी, Magyar, Bahasa Indonesia, Italiano, 日本語, 한국어, Norsk, Nederlands, Polski, Português, Română, Русский, Slovenčina, Svenska, ไทย, Türkçe, Українська, Tiếng Việt, 简体中文, 繁體中文
Pick by what you need
From VoiceInk's docs page Choosing Models, read October 9, 2026.
| Need | VoiceInk's starting point |
|---|---|
| Local dictation on Apple silicon | Parakeet V3 |
| English only | Parakeet V2 or Parakeet Unified |
| Other languages | Parakeet Ultra, Nemotron, Whisper, Cohere Transcribe or SenseVoice, checked against their language lists |
| Text appearing while you speak | A model that supports real-time transcription. Nemotron requires it |
| Apple's own recognition | Apple Speech, on macOS 26 or later |
| Audio or video files | A model that supports file transcription |
| Intel Mac | A cloud transcription provider |
Its docs add a caution worth repeating: a multilingual model does not support every language. Check the Mode's Language picker before you download.
The local models, and their languages
From VoiceInk's docs page Local Models, read October 9, 2026.
| Model | Languages | Real-time |
|---|---|---|
| Parakeet V2 | English | Optional |
| Parakeet V3 | English and supported European languages | Optional |
| Parakeet Unified | English | Optional |
| Nemotron Latin | English, French, German, Italian, Portuguese, Spanish | Required |
| Nemotron Multilingual | Several scripts, including Arabic, Chinese, Hindi, Japanese and Korean | Required |
| Whisper (Tiny to Large v3 Turbo) | Many | No. It processes after you stop |
| Cohere Transcribe | 14, including Arabic, Chinese, Japanese, Korean and Vietnamese | No |
| SenseVoice Small | English, Japanese, Korean, Cantonese, Mandarin | No |
You can also import a whisper.cpp model file. Larger Whisper models take more storage and longer to load.
Two models, not one
VoiceInk separates the transcription model, which turns audio into text, from an optional enhancement model, which cleans that text up. For local cleanup its docs point to VoiceInk Refine, which needs an Apple silicon Mac with at least 16 GB of memory and has a fixed cleanup job with no custom prompts. For custom prompts they point to a local model server or a hosted provider with your own API key. If dictation feels slow, turn enhancement off once to see which of the two models is the slow one. That tip is VoiceInk's own.
Test on your own voice
VoiceInk's docs say speed and accuracy vary with your microphone, language, recording and hardware, and they give you the tools to compare: History, Analyze, for a recording's timings, and Dashboard, View Insights, Model Performance, for summaries. A day with each candidate tells you more than any chart.
How YapToText handles the same choice
YapToText takes the opposite approach: fewer choices. It runs Whisper on the Mac, its App Store build ships with a speech model already inside, and a separate cleanup model runs locally too. It does not offer Parakeet, Nemotron or Cohere Transcribe. If you want to pick among engines, VoiceInk is the better fit. If you want to install and start talking, that is what mine is for. The local speech model comparison sets Whisper, Parakeet, Nemotron and Cohere Transcribe side by side from their model cards, and the Parakeet vs Whisper page goes deeper on those two.
This app is completely free. You don't have to make an account. There are no subscriptions, no difficult downloads, and nothing is locked. Made for Apple silicon Macs. Source on GitHub.
The short version
- Start with. Parakeet V3 on Apple silicon.
- English only. Parakeet V2 or Unified.
- Other languages. Check the language picker: Whisper covers the most.
- Live text. A real-time model such as Nemotron.
- Intel. A cloud provider.
If VoiceInk adds or changes a model, open an issue on GitHub and I will update this page. Every one gets read.
This app is completely free. You don't have to make an account. There are no subscriptions, no difficult downloads, and nothing is locked. Made for Apple silicon Macs. Source on GitHub.
Related
Sources
- https://tryvoiceink.com/docs/recommended-models (checked 2026-10-09)
- https://tryvoiceink.com/docs/local-models (checked 2026-10-09)
- https://tryvoiceink.com/docs/voiceink-refine (checked 2026-10-09)
- https://tryvoiceink.com/docs/realtime-transcription (checked 2026-10-09)
- https://github.com/ryleighnewman/YapToText (checked 2026-10-10)
- https://apps.apple.com/us/app/yaptotext/id6786382289 (checked 2026-10-10)
Questions
What is the best model in VoiceInk?
Its docs do not rank them. They name Parakeet V3 as the starting point for local dictation on Apple silicon, and it is the model onboarding uses.
Does VoiceInk support Parakeet?
Yes: Parakeet V2, V3, Ultra and Unified are in its local model catalog.
Which VoiceInk model supports my language?
Check the Language picker in a Mode. Parakeet V3 covers English and supported European languages, and Whisper covers many more.
Does VoiceInk work on an Intel Mac?
The app warns that local models do not work reliably on Intel and suggests a cloud transcription provider.
What is VoiceInk Refine?
A local text cleanup model, not a speech model. It needs an Apple silicon Mac with at least 16 GB of memory.