Pepys

Comparison

Local, Private and Offline Dictation Apps for Mac

Where each app processes your speech, what works offline, and what leaves your device, in each vendor's own words.

Yes. Several dictation apps process speech locally on your Mac and keep working without a connection. Pepys Dictate, Superwhisper, VoiceInk, MacWhisper and Whisper Notes all offer local workflows in different forms, while Wispr Flow uses cloud transcription. Local does not mean identical: check offline use, optional cloud features, accounts and model downloads. Facts were read on 8 October 2026.

Which dictation apps process speech locally?

Pepys Dictate, VoiceInk, FluidVoice, MacWhisper and Whisper Notes process speech on the device by default, according to their own pages. Superwhisper and Spokenly let you choose local or cloud models. Apple Dictation is on-device for some languages. Wispr Flow and Willow Voice transcribe in the cloud by default.

Every cell below was read from the vendor's own pages on 8 October 2026. "Not stated in public docs" means the pages read do not say; it is not a claim that the feature is missing. For how Apple's own Dictation behaves, see our guide to offline dictation on Mac.

Where speech and text are processed, offline use, model downloads, cloud options and accounts
ProductSpeech processingFormatting and text processingWorks offline?Models downloaded?Cloud option?Account required?
Pepys DictateLocal: speech recognition runs on your MacLocal: formatting runs on your MacYes, once its models are downloaded; a paid licence key is activated online onceYes: about 1.3 GB on first launch; Edit by voice adds a separate model fileNo cloud transcription. An optional advanced setting can send a transcript to an AI provider you configure; it is off by defaultNo account
VoiceInkLocal by default: its FAQ says transcription runs locally; optional cloud transcription uses your own API keyYour choice: VoiceInk Refine runs on-device; AI enhancement sends text to the provider you select; Ollama can keep it localYes after setup: its README says processing is 100% offline; cloud options need a connectionYes: downloaded under AI Models, Local; Apple Speech needs macOS 26 or laterYes, optional: cloud transcription and AI enhancement with your own API keyNot stated in public docs
FluidVoiceLocal: its README says speech recognition runs on your MacOptional: a local runtime (Fluid Intelligence, not open source) or cloud AI providers with your own API keyNot stated in public docsYes: speech models are about 250 MB to 2.9 GB, except Apple Speech; the optional local runtime is about 3.5 GBYes, optional: cloud AI providers for enhancement with your own API keyNot stated in public docs
SpokenlyYour choice: local models keep speech on the device; built-in cloud models send audio to its backend, which forwards itAI Instructions can send transcription and optional context to AI services, even after local transcriptionYes with local models, per its pricing pageNot stated in public docsYes: built-in cloud models, or your own API keysNot needed for local models, per its pricing page
SuperwhisperYour choice per mode: local voice models or cloud modelsYour choice per mode: a local language model (S1-mini), a cloud language model, or no AI processingYes with local models: its model page says local models work offline; cloud models need a connectionYes for local models: one-time downloads of 75 MB to 3 GB, per its docsYes: cloud voice and language models (Pro), or your own keysNot stated in public docs
MacWhisperLocal: its homepage says it transcribes locally on your Mac with local AI modelsOptional AI clean-up uses your own AI key (cloud providers) or a local model such as Ollama or LM StudioYes: its homepage lists an Offline feature for local processingNot stated in public docsYes, optional: remote models and providers with your own API keysNot stated in public docs
Whisper NotesLocal: its site says transcription runs on the device and audio never leaves itNot stated in public docsYes: its privacy page says no internet connection is requiredNot stated in public docsNone statedNo account, per its FAQ
Apple DictationOn-device for some languages: Keyboard settings show whether it is processed on your device; otherwise audio goes to Apple serversNot stated in public docsSome languages: Apple's macOS 13 and 14 pages say no internet on Apple silicon for supported languages; its current pages do not say soYes: Apple's Feature Availability page says on-device Dictation requires a download of speech modelsYes when not on-device: dictated content is processed on Apple serversNot stated in public docs
Wispr FlowCloud: its data-controls page says transcription always occurs on the cloudCloud service: it uses the app name and surrounding text for formatting and punctuationNo: transcription always occurs on the cloud, per its data-controls pageNo local modelsYes: cloud onlySign-in is a setup step, per its setup guide
Willow VoiceCloud: its help center says Willow uses cloud-based AI modelsCloud models; optional Context Awareness reads small bits of text around the cursorPaid plans: its help center lists offline dictation under the Individual Pro plan, and its homepage says it is powered entirely by your deviceNot stated in public docsYes: cloud models by defaultSign-in required, per its help center

Private, local and offline are not the same thing

These words are often used as if they meant the same thing. Each answers a different question. A local app can still make network requests for licensing, updates or analytics, and a cloud service can still have strong security controls. Our guide to local vs cloud dictation covers the basics.

  • Local processing: speech or text is processed on your own device.
  • Offline: the relevant function keeps working without an internet connection, after any required models or resources are on the device.
  • Private: a broader idea that covers processing, storage, telemetry, account data and vendor practices, not only where speech is recognised.
  • Cloud processing: speech or text is processed on remote infrastructure.

Five kinds of dictation architecture

Grouping by architecture is more useful than ranking by popularity. Each group is assigned from what the vendor's own pages state, and an app can sit in a group while offering options from another.

Dictation apps grouped by architecture
TypeHow it worksApps on this page
Local-first dedicated dictationSpeech recognition runs on the device by default, with cloud features optional or absent.Pepys Dictate, VoiceInk, FluidVoice
Hybrid and model-selectableYou choose local or cloud models, so the architecture depends on your settings.Spokenly, Superwhisper
Local transcription suites with dictationBuilt mainly for files and recordings, with local processing and a dictation feature.MacWhisper, Whisper Notes
Built-inPart of the operating system.Apple Dictation
Cloud and cross-platform dictationSpeech recognition runs on the vendor's servers by default.Wispr Flow, Willow Voice

Platforms and price model for each app

Platform and price-model facts decide whether a local app can replace a cloud one for you. Prices and licence terms for the one-time options are on our one-time purchase page.

Platforms, one-time option and main trade-off for each app
ProductMacWindowsiOSAndroidOne-time optionMain trade-off
Pepys DictateYes (Apple silicon, macOS 13 or later)No (planned, not available)No (an iOS keyboard is in development)No (planned, not available)YesMac only today, one Mac per licence, and a model download on first launch
VoiceInkYesNoBeta (a separate iOS download is linked)NoYesOptional cloud and AI enhancement send audio or text to the provider you choose; Apple silicon is recommended
FluidVoiceYes (macOS 15 or later; Apple silicon for most models)Not stated in public docsNot stated in public docsNot stated in public docsFree and open source (GPLv3)Anonymous analytics are on by default, and it needs macOS 15 or later
SpokenlyYesYesYesNo (not yet available, per its Android page)Not stated in public docs; local models are free and Pro is a yearly subscriptionAI Instructions can send text out even when transcription ran locally
SuperwhisperYesYesYesYesYes (Lifetime)Staying local depends on each mode's settings, and some local and all cloud models need Pro
MacWhisperYesNot stated in public docsYes (its FAQ lists an iOS and iPad version)Not stated in public docsYes (Pro)System-wide dictation is free but only in the direct-download build, and its clean-up and AI prompts need Pro and an AI service you configure, and its privacy policy is dated February 2024
Whisper NotesYes (Apple silicon, macOS 14 or later)Not stated in public docsYes (iPhone and iPad)Not stated in public docsYesSystem-wide voice typing on the Mac is only in the direct-download build, not the Mac App Store build
Apple DictationYesNo (Apple devices only)YesNo (Apple devices only)Not applicable: built into macOSOn-device only for some languages; check Keyboard settings for yours
Wispr FlowYesYesYesYesNo: its pricing page lists only recurring paid plansCloud transcription needs a connection; its data controls decide storage and training
Willow VoiceYesYesYesNo (coming soon)No: its pricing page lists recurring plansCloud by default, and offline dictation is a paid-plan feature

How we compared them

  • Where speech recognition runs, and where formatting or AI processing runs, in the vendor's own words.
  • Whether it works offline, what must be downloaded first, and what still needs a connection.
  • What leaves the device: audio, text, context, analytics, account and licensing data, as each vendor states it.
  • Platforms and price model, so a local app can be weighed against a cloud one.
  • Wording: 'private' means only what a vendor's pages say, and silence is shown as 'Not stated in public docs'.

Pepys Dictate is our app. Every product here was read from its vendor's own pages on 8 October 2026. Entries are grouped by architecture, not ranked, and no app gets a privacy score or a secure label. Every product lists trade-offs, ours included.

Which dictation apps actually work offline?

It depends on the app and on your settings. Offline here means speech recognition keeps working without a connection, after any required models are downloaded. The answer differs for formatting, cloud models and licence checks, so each is stated separately.

  • Pepys Dictate: speech recognition and formatting both run on your Mac, so dictation works offline once the models are downloaded on first launch. A paid licence key is activated online once, and a licence check runs at launch when online.
  • VoiceInk: local transcription runs on your Mac after setup, and its README says processing is 100% offline. Cloud transcription and AI enhancement through a provider need a connection, and its pages do not detail licence activation.
  • Superwhisper: its model page says local models work offline, after a one-time download. Cloud models need a connection, so a mode that uses one does not work offline. A fully local setup pairs a local voice model with a local language model.
  • MacWhisper: its homepage lists an Offline feature for local processing. Remote models and providers need a connection, and system-wide dictation is free but only in the direct-download build, with clean-up and AI prompts in Pro.
  • Whisper Notes: its privacy page says no internet connection is required. It does not describe formatting or licence checks.
  • Spokenly: local models work offline, per its pricing page. Built-in cloud models and AI Instructions need a connection.
  • FluidVoice: speech models run on the device after download. Its README does not explicitly state offline use.
  • Apple Dictation: only for some languages. Apple's macOS 13 and 14 pages say no internet is needed on Apple silicon for supported languages, and its current pages do not say so.
  • Wispr Flow: no. Its data-controls page says transcription always occurs on the cloud.
  • Willow Voice: its help center lists offline dictation under the Individual Pro plan. Otherwise it uses cloud models.

What actually leaves your device?

This table reports what each vendor's own pages say about each kind of data. It does not rate any of them. Where the pages are silent the cell says so, because silence is not the same as nothing leaving the device.

What each app's own pages say about audio, text, context, analytics, licensing and history
ProductRaw audioRecognized textContext and promptsAnalytics and telemetryAccount and licensingHistory and storage
Pepys DictateStays on your Mac, per its privacy page: recordings are never uploadedTranscripts and formatted text stay in a local history database; the optional AI-provider setting can send a transcript, and is off by defaultSelected text, spoken edit instructions and transcribed files are not uploaded. With the optional, off-by-default "Use words on my screen", the focused field's text and the window title are read on the Mac for one dictation and not uploadedOpt-in only: content-free events such as event names, app version and macOS version, and "a step failed" with one short error code (never a message, text or file name). Off until you agreeNo account. Activating a licence key sends the key, your Mac's name and its organisation id to its payment provider, and a check runs at launch when onlineA local database on your Mac, kept on the schedule you set
VoiceInkLocal by default. If you choose cloud transcription, audio goes to the provider you selectedLocal by default. AI enhancement sends transcription text to the provider you selectVocabulary and any enabled context go with AI enhancement; VoiceInk Refine cleans text locally with no custom contextIts privacy docs describe no telemetry; they say the dashboard keeps text-free session metricsAccount not mentioned; its licence-activation network use is not detailedTranscripts, saved audio and AI request metadata are stored locally, with retention set in History Settings
FluidVoiceOn-device by default; a cloud AI provider receives data only if you opt inOn-device by default; enhancement text goes to the provider you add if you opt inNot stated in public docsAnonymous analytics are on by default and can be disabled; a weekly batch with a random install ID, date, app version and macOS label is still sentFree, with no licence keyNot stated in public docs
SpokenlyLocal models: never leave the device, per its privacy policy. Built-in cloud models: sent to its backend, which says it does not store them. Your own key: sent straight to your providerAI Instructions can send transcription text to AI servicesOptional context for a Mode may go to AI services with AI InstructionsNot stated in public docsEmail sign-in is used for cloud models and Pro in its Chrome extension; the desktop apps' account requirement is not statedNot stated in public docs
SuperwhisperA fully local setup keeps audio on the device. Cloud models process audio through API callsA fully local setup keeps text on the device. Cloud language models receive text through API callsCustom vocabulary words are sent to a cloud voice model as recognition hints; with a local voice model they stay on the deviceNot stated in public docsNot stated in public docsAll transcription history is stored locally; FileSync syncs modes and settings, not audio or transcripts
MacWhisperIts privacy policy says no audio, text or other data leaves the device; its homepage also lists optional remote providersAs above: the policy says no text leaves the device; AI features can use providers you addNot stated in public docsIts policy mentions anonymous analytics events about licensing status and does not say what they containAccount not mentioned; the policy describes no separate licence check beyond those eventsNot stated in public docs
Whisper NotesNever uploaded to its servers, per its privacy pageTranscripts stay on the device, per its privacy pageNot stated in public docsIts privacy page says it uses no analytics or tracking toolsNo account; whether a licence check makes network requests is not statedStays on the device unless you choose to share it
Apple DictationOn-device when Keyboard settings say so; otherwise sent to Apple serversApple says server-processed dictation is not stored unless you opt in to Improve Siri and Dictation, and that request history including transcripts may be kept up to two years under a random identifierNot stated in public docsNot stated in public docsBuilt into macOSA separate Siri and Dictation history is stored on the device and can be deleted in Settings
Wispr FlowProcessed in the cloud; encrypted in transit and at rest, per its pageProcessed in the cloud; the Dictation Cloud Storage setting controls whether transcripts, audio and history are stored on its serversWith Context Awareness on, text from the active app window may be used to improve accuracy; you can turn it offUsage statistics such as word counts are collected regardless of data controlsSign-in required during setupOptional cloud storage of transcripts, audio and dictation history
Willow VoiceIn Private Mode, its default, no audio is stored on its serversIn Private Mode, no dictated text is collected; the optional Help Improve setting collects recognized text in anonymized formOptional Context Awareness text is not stored or linked to your account, per its help centerNot stated in public docsSign-in requiredTranscript history is stored only on your device
  1. Pepys Dictate(our app)

    Best for: Mac users who want local-first dictation with no account

    Pepys Dictate runs speech recognition and formatting on your Mac, with no account and no cloud transcription. It is $9.99 once, plus tax where applicable, for one Mac, with future Mac app updates included.

    Pros

    • Its privacy page lists what leaves the Mac: model downloads, licence activation, update checks and opt-in content-free events.
    • Works offline once its models are downloaded on first launch.
    • A free plan of 2,000 words a week and a 7-day trial of everything, so you can test it before you buy.
    • Selected text, spoken edit instructions and transcribed files are not uploaded.

    Cons

    • Mac only today, and one Mac per licence. An iOS keyboard is in development, Windows and Android are planned, none is available, and their licensing has not been announced.
    • Needs an Apple silicon Mac on macOS 13 or later, and a model download on first launch.
    • Not open source, so you cannot read the code to confirm what its privacy page says.
    • No meeting notes, team features, compliance features or certifications.
  2. VoiceInk

    Best for: People who want open-source code and local transcription by default

    VoiceInk's FAQ says transcription runs locally on the Mac by default, and its README says processing is 100% offline. Cloud transcription and AI enhancement are optional and use a provider you choose.

    Pros

    • Source code is public on GitHub under GPL v3, so you can read it or build it yourself.
    • Its privacy docs say transcripts, saved audio and AI request metadata are stored locally, with retention set in History Settings.
    • VoiceInk Refine cleans text on-device, and Ollama can keep AI enhancement on your Mac.

    Cons

    • AI enhancement sends transcription text, vocabulary and any enabled context to the provider you select.
    • Its docs warn that a locally launched command-line tool is not necessarily an offline model.
    • Apple silicon is recommended, and local models do not work reliably on Intel Macs.
    • Its pages do not detail licence-activation network use, and they do not mention accounts.

    Pepys Dictate compared with VoiceInk

  3. FluidVoice

    Best for: A free, open-source app with on-device recognition by default

    FluidVoice's README says speech recognition runs on your Mac, and that your voice and text stay on the machine unless you opt in to a cloud AI provider. It is free and open source under GPLv3.

    Pros

    • Speech models range from about 250 MB to 2.9 GB, and Apple Speech needs no download.
    • API keys for optional cloud providers are stored in the macOS Keychain, per its README.
    • Enhancement can use a local runtime instead of a cloud provider.

    Cons

    • Anonymous analytics are on by default, and a weekly batch with a random install ID is still sent when they are disabled.
    • Its local runtime, Fluid Intelligence, is private and not open source.
    • Needs macOS 15 or later, and Apple silicon for every model except Whisper.
    • Its README does not explicitly state offline use.

    Pepys Dictate compared with FluidVoice

  4. Spokenly

    Best for: People who want free local models and the option of cloud ones

    Spokenly's privacy policy says audio recordings and transcriptions never leave the device with local models. Built-in cloud models send audio to its backend, which says it does not store the recordings.

    Pros

    • The free local models need no account, per its pricing page.
    • Its policy says it keeps no permanent copy of recordings on its servers.
    • With your own API key, data goes directly to your provider and not through its servers.

    Cons

    • AI Instructions can send the transcription and optional context to AI services, even after local transcription.
    • Its pages do not state whether the desktop apps need an account.
    • Its pricing page lists a yearly Pro plan and free local models, and states no one-time option.

    Pepys Dictate compared with Spokenly

  5. Superwhisper

    Best for: People who want to choose local or cloud models for each task

    Superwhisper's docs say that in a fully local setup no audio or text leaves your machine. Each dictation has a voice-to-text stage and an optional language-model stage, and each can be local or cloud, set per mode.

    Pros

    • Local voice models work offline after a one-time download, per its model page.
    • It says it does not train AI models on your audio or transcribed text, and that history is stored locally.
    • With a local voice model, custom vocabulary stays on the device.

    Cons

    • Staying local depends on the model settings of every mode.
    • Custom vocabulary words are sent to a cloud voice model as recognition hints.
    • Some local models and all cloud models need Pro.

    Pepys Dictate compared with Superwhisper

  6. MacWhisper

    Best for: People who mainly transcribe files and meetings, with some dictation

    MacWhisper's homepage says it transcribes locally on your Mac with local AI models, and lists an Offline feature. Its privacy policy says no audio, text or other data leaves the device.

    Pros

    • Built for local transcription of files, meetings and system audio, with dictation as one feature.
    • AI features can use a local model such as Ollama or LM Studio, or your own provider keys.

    Cons

    • System-wide dictation is only in the direct-download version and needs a Pro licence, per its docs.
    • Its privacy policy is dated February 14, 2024, and mentions anonymous licensing analytics events without saying what they contain.
    • Its homepage lists optional remote providers, so what leaves the Mac depends on the models you pick.

    Pepys Dictate compared with MacWhisper

  7. Whisper Notes

    Best for: iPhone and iPad users who want local transcription, with an optional Mac build

    Whisper Notes says transcription runs on the device, audio never leaves it, and no account or internet connection is needed. Its privacy page says it uses no analytics or tracking tools.

    Pros

    • No account, per its FAQ, and no internet connection required, per its privacy page.
    • Runs on iPhone, iPad and the Mac.
    • Its Mac direct build includes a free weekly word allowance, per its site.

    Cons

    • System-wide voice typing on the Mac is only in the direct-download build, not the Mac App Store build.
    • Whether a licence check makes network requests is not stated.
    • Its pages do not describe formatting or clean-up features.
  8. Apple Dictation

    Best for: Nothing to install, if Keyboard settings show on-device Dictation for your language

    Apple says Keyboard settings show whether Dictation is processed on your device. When it is, audio and transcripts are not sent to Apple servers. Otherwise dictated content is processed on Apple servers.

    Pros

    • Built into macOS, so there is nothing to install.
    • Apple says server-processed dictation is not stored unless you opt in to Improve Siri and Dictation.
    • A separate history stays on the device and can be deleted in Settings.

    Cons

    • On-device processing covers only some languages, so check Keyboard settings for yours.
    • Apple's privacy page says request history, including transcripts, may be kept up to two years under a random identifier, and a reviewed subset longer.
    • Its macOS 13 and 14 pages say no internet is needed on Apple silicon for supported languages, and its current pages do not say so.

    Pepys Dictate compared with Apple Dictation

What local processing costs

Running models on your own machine moves the work from a vendor's servers to your Mac. The pages read give some numbers, and this page makes no claim about speed, CPU, memory or battery use, because none has been measured here.

  • Storage: local models are downloads. Pepys Dictate's models are about 1.3 GB on first launch. Superwhisper lists local models from 75 MB to 3 GB, and FluidVoice lists speech models of about 250 MB to 2.9 GB.
  • Apple silicon: Pepys Dictate needs it, and so do Whisper Notes and most FluidVoice models. VoiceInk recommends it and warns that local models do not work reliably on Intel Macs.
  • Memory: Superwhisper's model page gives no specific RAM requirement, and several vendors' pages state none.
  • Older machines: Superwhisper's page recommends cloud models for older machines, since they are fast on any hardware.

When cloud dictation makes more sense

Cloud transcription is not a worse design, just a different one. A cloud service can do things a one-device local app does not, and the pages read show real examples.

  • Cross-device continuity: Wispr Flow's pricing page lists dictation on Mac, Windows, iOS and Android on one plan.
  • Lower local hardware needs: cloud models need no download and run on any machine, which is why Superwhisper recommends them for older hardware.
  • Hosted history and sync: Wispr Flow's Dictation Cloud Storage setting can keep transcripts, audio and history on its servers.
  • Teams and enterprise administration: Willow Voice's Enterprise plans include SSO and SAML, a signed MSA and DPA, and enforced privacy settings, per its help center.
  • Managed models and updates: the vendor improves the model on its servers, with nothing for you to download.
  • Mobile: most cloud and cross-platform apps here have iPhone apps today. See the Wispr Flow pricing page for one example.

When local-first dictation makes sense

Local-first dictation tends to suit you if most of these are true.

  • You prefer your speech to be recognised on your own device.
  • Your internet is unreliable, or you dictate on planes and while travelling.
  • You work on sensitive material and want fewer places it can travel.
  • You dictate mainly on one Mac.
  • You would rather depend less on a vendor's servers and plans.
  • You are comfortable with model downloads and a capable Mac.

Which kind of dictation should you choose?

Choose by architecture and need, not by which app is most popular.

  • Choose local-first dictation if you want recognition on your own device by default, mostly on one Mac, with few network calls. For a Mac-only comparison grouped by need, see our Mac dictation app comparison.
  • Choose cloud-connected dictation if you need Mac, Windows and mobile apps now, history across devices, or team administration.
  • Choose hybrid or model-selectable tools if you want to pick local or cloud per task, and will check each mode's settings. Compare Superwhisper and the other apps in our roundup of dictation apps.

What is the best private transcription software?

For recordings you would rather not upload, the most private option is software that transcribes files on your own computer. Pepys Dictate does this on a Mac: drop in a file or record a voice memo and it is transcribed there, not uploaded, up to 10 minutes per file on the free plan and any length with Pro.

Pepys Transcribe, our separate web app, is cloud transcription: it does not process on your device. Local and cloud transcription are different privacy models, so compare them on the points below. For what to ask any service, see our guide to confidential transcription.

Local file transcription and cloud transcription compared on where audio goes, retention, training and control
QuestionLocal file transcription (Pepys Dictate)Cloud transcription (Pepys Transcribe)
Where the audio goesStays on the Mac, where speech recognition runsUploaded to cloud storage and sent to a third-party speech-to-text provider
How long it is keptTranscripts stay in a local history on the Mac until you clear itThe original recording is deleted automatically about 30 days after upload, or sooner if you delete it; the transcript stays until you delete it
TrainingFiles and transcripts are not uploadedPepys does not use your content to train its own models; provider controls vary by service
Network useOnline to download models, activate a license and check for updates; content-free usage events only if you opt in; an optional setting, off by default, can send a transcript to an AI provider you configureNeeds a connection to upload and transcribe
Deletion and controlClear History in the app, or delete the app and its folderErase the recording from the editor's player, delete the transcript, or delete your account

Where Pepys Dictate fits

Pepys Dictate is local-first AI dictation software for Mac. Its privacy page says speech recognition and formatting run on your Mac, with no account and no cloud transcription, and that audio is never uploaded. The app goes online to download its models, activate a licence key, check the licence at launch when online, and check for updates. Anonymous usage events are opt-in.

An optional advanced setting can send a transcript to an AI provider you configure. It is off by default. Pepys Dictate does not claim that nothing ever crosses the network: the list above is what it sends.

Pepys Dictate is a one-time purchase with no recurring subscription: $9.99 once for one Mac, plus tax where applicable, with future Mac app updates included. A free plan covers 2,000 words a week.

Pepys Dictate is Mac only today: Apple silicon, macOS 13 or later, one Mac per licence. An iOS keyboard is in development, and Windows and Android are planned. None is available, and the Mac licence does not cover them. Its history lives on one Mac, so it does not follow you across devices. It has no meeting notes, team administration or certifications.

  • It suits you if you want local-first processing, no account and a one-time price on a single Apple silicon Mac.
  • It does not suit you if you need Windows, iPhone or Android today, history across devices, or team controls.
  • Try Pepys Dictate for Mac, or compare it with Wispr Flow.

Questions about these apps

What is the best private dictation app?

There is no single best one. If you want speech recognized on your Mac and dictate on one Mac, Pepys Dictate, VoiceInk, FluidVoice, MacWhisper and Whisper Notes process speech locally by default, per their own pages. If you want to choose local or cloud per task, Superwhisper and Spokenly allow it. If you need Windows, iPhone and Android today, cloud-connected apps such as Wispr Flow cover more platforms. Check optional cloud features, accounts and analytics for each.

Does Pepys Dictate work offline?

Yes, once its models are downloaded. Pepys Dictate downloads its speech and formatting models on first launch, about 1.3 GB. Activating a paid licence key needs a connection once, and its privacy page says the app works offline after that. A licence check runs at launch when online.

Does Pepys Dictate upload audio?

No. Pepys Dictate's privacy page says recordings are processed on your Mac and never uploaded, and that transcripts and formatted text stay in a local history database. The app goes online to download models, activate a licence, check for updates and, only if you opt in, send content-free usage events. An optional advanced setting can send a transcript to an AI provider you configure. It is off by default.

Does Superwhisper work offline?

Yes, with local models. Superwhisper's model page says local models work offline and keep audio on your device, after a one-time download of 75 MB to 3 GB. Cloud models need a connection, so a mode that uses one does not work offline. Its docs describe a device-only setup of a local voice model plus a local language model, in which no audio or text leaves the machine.

Is Wispr Flow local or cloud based?

Cloud based. Wispr Flow's data-controls page says transcription always occurs on the cloud. Its Dictation Cloud Storage setting controls whether transcripts, audio and history are stored on its servers, and its pages describe no local or on-device option. Its pricing page lists dictation on Mac, Windows, iOS and Android.

Is local dictation more private?

Not automatically. Local processing keeps speech recognition on your device, which avoids sending audio to a vendor's servers. Privacy also depends on optional cloud features, analytics, licensing checks, account data and what the app stores. A cloud service can have strong security controls and clear retention terms. Compare each app's own data-flow statements, as in the table above.

What is the difference between local and cloud dictation?

Local dictation runs the speech recognition model on your own device, so audio can become text without leaving it. Cloud dictation sends audio to remote servers that return text. Local needs model downloads, storage and a capable Mac. Cloud needs a connection but little local hardware, and can sync across devices. Many apps offer both.

Can speech-to-text run entirely on a Mac?

Yes. Pepys Dictate, VoiceInk, MacWhisper, Whisper Notes and Superwhisper with local models run speech recognition on the Mac after downloading models, per their own pages. How entirely depends on settings: optional cloud models, AI clean-up through a provider, licence checks and update checks can still use the network. Apple Dictation runs on-device only for some languages.

Where these facts came from

Want local-first dictation on your Mac?

Try Pepys Dictate for Mac: $9.99 once, plus tax where applicable, with future Mac app updates included. Speech recognition and formatting run on your Mac, and it works offline once its models are downloaded.

Apple silicon, macOS 13 or later. More about Pepys Dictate