Superwhisper

Superwhisper is a dictation app that types your speech into any application on macOS, Windows, or iPhone. It runs Whisper-class models on your own machine by default, so a full drafting session works on a plane with no wifi, and it still sells a one-time lifetime licence in a category that has otherwise gone subscription-only.

What You Get#

• System-wide dictation: Hold a shortcut, speak, release — formatted text lands in whatever app has focus: Scrivener, Word, Google Docs, email. No per-app integration.
• Custom modes per app: Formatting rules that switch by destination — the vendor ships a fiction mode that leaves dialogue alone and a newsletter mode that adds headers.
• Super Mode: An LLM pass that punctuates, breaks paragraphs and trims false starts from a raw spoken draft, running on local models (Llama 3.2 3B up to GPT OSS 20B) or cloud ones.
• Vocabulary and Replacements: Two mechanisms for invented names. Vocabulary passes words to the model as hints, though the docs warn that overloading it "can confuse the AI transcription model." Replacements is a post-transcription find-and-replace — the one that reliably sticks.
• File transcription: Existing audio or video in, text out. Pro only.

Who It's For#

• Authors who dictate first drafts: Speech runs near 150 words a minute against 40–50 typed, and Super Mode is built for the "get it down, shape it later" workflow.
• Writers with RSI or hand fatigue: A full-time typing replacement, not a novelty.
• Anyone with an unpublished manuscript: The local path covers transcription and AI cleanup end to end.

Pricing: Free forever on the three smallest on-device Whisper models (Fast, Nano and Standard, 75–500 MB), with unlimited words, 100+ languages and no expiry. Pro at $8.49/month, $84.99/year, or $249.99 lifetime adds the larger local models, every cloud model, file transcription and bring-your-own-API-key. A 40% student discount and 30-day refund apply, and one licence covers all three platforms.

Strengths#

• The lifetime licence: $249.99 once beats Wispr Flow's $144/year inside two years, and nothing else in this class still offers one.
• Offline genuinely works: "The audio is processed on your own hardware and never touches the internet." Wispr Flow and Aqua Voice are cloud-only and stop when the connection does.
• A free tier, not a trial: It "never expires" and does not meter words. Wispr Flow caps you at 2,000 words a week on desktop, Willow Voice likewise.
• Actively built: 4.4/5 from 812 App Store ratings, several releases a month through August 2026, by a bootstrapped six-person team in Toronto.

Limitations#

• "On-device" is a setting now, not a guarantee. Pro added cloud transcription and cloud LLMs, where "audio goes to the Superwhisper proxy, gets transcribed, and comes back." The privacy policy never mentions that pathway — it says "there is no need for data transmission." The local option is real, but selecting it is on you.
• Windows lags Mac badly: macOS is at 2.18.0, Windows at 1.5.10 — separate version lines, not a point release apart. Windows only went generally available around November 2025.
• Intel Macs are throttled, not excluded: the largest local models effectively want Apple Silicon.
• No Android, Linux or web version, and no SRT/VTT subtitle export — a gap conceded to MacWhisper on Superwhisper's own comparison page.
• Accuracy claims are unverified: no independent benchmark covers Superwhisper specifically, and Dragon still leads on dense technical vocabulary. "100+ languages" is an aggregate across the model lineup, not a promise from any one model.

Alternatives#

MacWhisper is better value on Mac alone — about €65 once, with free-build dictation and the subtitle export Superwhisper lacks — but no Windows version. Wispr Flow is the only one covering Android, at $15/month with no offline mode. Dragon Professional Individual suits only authors with an established voice profile or heavy specialized vocabulary.