Typeless Alternative: Hold Fn, Speak. Your Voice Stays Local.

Typeless is good software, and the flow is the reason people stay with it: hold Fn, speak, release, and what lands at the cursor is already tidy — the "um"s gone, the sentences straightened, the tone matched to the app you are typing into. For a Slack message that is usually the better output.

Three things about that flow are worth knowing before you rent it by the month.

1. The recording leaves the machine

Every press of Fn sends your audio to a server. Typeless states a zero-retention policy and we have no reason to doubt it — but a retention policy is a promise about what a company does with the file, not a statement that the file was never there. For a privileged conversation, a patient note or an unreleased product decision, that difference is the whole question.

Whisper Notes runs its models on your own Mac: Parakeet V3 by default, Whisper Large V3 Turbo when you need the wider language list, and SenseVoice when you select it for Chinese, Japanese, Korean or Cantonese. There is no upload path to trust, which you can check in about thirty seconds by turning on airplane mode and dictating a sentence.

2. Cloud polish costs a round trip

The Typeless path is speech recognition, then an upload, then the LLM rewrite, then the text coming back. Each step is fast; four of them in sequence are not.

Both numbers above are our own runs on an M4 Pro, and a 20-second clip is mostly start-up rather than throughput. On longer audio the throughput shows: Parakeet V3, the default engine, gets through roughly 60 times realtime on Apple Silicon — a 17.5-minute meeting finished in 16.7 seconds on our M5 Air. Switch to the Whisper route, Large V3 Turbo with its wider language list, and expect roughly 11 times realtime instead.

The trade in one line: Typeless waits on a network to hand you polished prose; Whisper Notes stays local and hands you the recording, tidied but not rewritten.

3. It is a subscription

Typeless runs cloud infrastructure — servers and third-party LLM APIs — and that costs money every month the service stays up. A subscription is the correct way to price it: the cost is recurring, so the price is too.

Whisper Notes has no backend, so the compute is your Mac's and the bill is one payment. That is not a better business model, it is a different one, and it is only cheaper if you keep the app for more than a few months.

Both apps do system-wide dictation on the Fn key, and that is where the overlap ends.

Audio file import: Drag in a recorded audio or video file — an interview, a meeting, a lecture — and Whisper Notes transcribes it. Typeless is built for live dictation and does not do this.

A choice of engine: Parakeet V3 is the default; on our own runs it is about five times faster than Whisper Large V3 Turbo on the same file, and it covers 25 European languages. Whisper Large V3 Turbo is the route to roughly 100 languages. SenseVoice is the one you select for Chinese, Japanese, Korean and Cantonese — it is a selection, not automatic routing. Typeless uses its own cloud models and there is nothing to pick.

Export formats: Text, SRT, VTT and JSON, so subtitles and structured output are covered. Typeless optimizes for one thing: text at the cursor.

If live dictation into Slack and email is the entire job, both apps do it and Typeless does the polishing part better. If files also pile up — recordings to transcribe, subtitles to produce, a model to choose per language — that is the part Whisper Notes was built around.

For some workflows Typeless is simply the better-fitting tool, and it is worth being specific about which ones.

Prose out of rough speech: If most of what you dictate is Slack messages and quick email, the cloud rewrite is worth the round trip. Say "so, um, like, I think we should maybe try this" and Typeless hands back "I think we should try this."

Windows: Whisper Notes is Mac and iOS only. On Windows, Typeless is one of the few dictation tools of this quality that covers both platforms, and no amount of preference for local processing changes that.

Formatting that follows the app: Typeless adjusts tone and structure to wherever you are typing — casual in Slack, more formal in Gmail. Doing that well takes a model larger than a laptop holds.

A free tier: Typeless gives 4,000 words a week at no cost, which is enough for light use. Whisper Notes gives 5,000 words a week on the Mac and then costs $14 once.

If you are comfortable with audio going to a server, and rewritten output is what you want from dictation, Typeless is the better-fitting tool and cross-platform besides. Whisper Notes is the one to buy when the audio has to stay on the machine, or when dictation is only part of what you need from the app.

Whisper Notes transcribes what you said. Typeless rewrites it, and for a message to a colleague the rewrite is usually what you wanted.

There is a cleanup pass on our side too, and it is worth being exact about where it stops. On the Mac direct download build, Gemma 4 runs on your machine: it removes filler words, fixes punctuation and gives the recording a title. It does not restructure your sentences and it does not change your register.

That boundary is deliberate rather than a missing feature. A model small enough to run on your laptop will not rewrite prose as well as a cloud pipeline built for it, and a confident rewrite of a clinical note is worse than an "um" left in.

So the transcript reads cleanly and still matches the recording. Those are two different products, and picking between them is easy once you know which output you need.

Choose Typeless if polished prose out of rough speech is the point of the tool, or you work on Windows.

Choose Whisper Notes if the transcript has to match the recording — depositions, clinical notes, interviews — or the audio must not leave the machine.