Skip to content

superwhisper vs VoiceInk, and where Wavenote fits

Both of these run speech models on your own Mac, both now run Parakeet, and both are what people recommend when someone asks how to dictate without sending audio anywhere. What separates them is what you pay and whether you can read the code.

The honest read on superwhisper

superwhisper is the more finished of the 2 and the one with a company behind it. Several model sizes to choose between, your own API keys for cloud models when accuracy matters more than privacy, meeting recording and summaries, Windows and iOS builds, and a lifetime licence for people who will not rent software. If you want to install one thing and have it work on the first try, this is the safer pick.

superwhisper's share card: the app's white triangular ribbon mark and the superwhisper wordmark on a purple to near-black gradient

The honest read on VoiceInk

VoiceInk is GPL v3, built in public on GitHub, and free forever if you build it in Xcode yourself. Its Modes system swaps an entire configuration per app or per URL, which is more than superwhisper's per-app handling does. It runs Parakeet and WhisperKit as well as whisper.cpp, it ships an iOS app, and the paid tiers are $29 to $69 once by how many Macs you want. If reading the source is the point for you, nothing else on this page competes.

The VoiceInk dashboard on macOS, with a left sidebar listing Dashboard, Record Audio, History, AI Models, Power Mode, Permissions, Audio Input, Dictionary, VoiceInk Pro, Settings and About, and panels reading You are 3.7x Faster with VoiceInk, 2h 48m 35s speaking time against 10h 18m 24s typing time, 24736 words captured, 146.7 average words per minute, and a 30-day trend chart

What actually decides it

Recognition quality is no longer the tiebreak, because both can run the same Parakeet model on the same Neural Engine. What is left is whether being able to audit and build the thing is worth more to you than superwhisper's polish, its meeting capture and its Windows build.

Side by side

superwhisperVoiceInkWavenote
Source availableNoYes, GPL v3No
Runs Parakeet on deviceYesYesYes
Per-app configurationYesYes, full ModesRewriting instruction only
Cloud fallback on your keyYesYesNo
JapaneseMultilingual modelMultilingual modelIts own Japanese model
Cost to dictate$8.49 a month or $249.99 once$29 to $69 once, or free from source$4.90 once for 2 devices

Where Wavenote fits

Wavenote is the smallest of the 3 and the least adjustable: 10.9 MB, a single model per language, one key to hold, no modes and no model picker, $4.90 once across 2 devices. Japanese runs on a model built for Japanese rather than on a multilingual checkpoint, which is the one place it is doing something neither of these does. It is also closed source, so if that is what brought you to this page, VoiceInk is your answer and Wavenote is not.

Which one to pick

superwhisper
You want local recognition with room to tune it, plus meeting capture and a Windows or iOS build.
VoiceInk
You want to read the source, build it yourself, or run a full per-app Modes setup.
Wavenote
You want one key and nothing else, you dictate in Japanese, or 2 devices for $4.90 matters more to you than a settings screen.

superwhisper pricing: Free tier, $8.49 a month, $84.99 a year, or $249.99 once. Checked against superwhisper.com on 7 August 2026.

VoiceInk pricing: One-time, $29 for one Mac up to $69 for three, or free if you build it yourself. Checked against tryvoiceink.com on 7 August 2026.