Mac dictation apps compared
Everything you want from Wispr Flow. Free.
Hush is free, recognizes your speech on your own Mac and is open source. Here is how it stacks up against Wispr Flow, Superwhisper and the rest, including what it doesn't do yet.
Free
No subscription, no word limit, no trial. Every feature is in the free download.
Private
Speech is recognized on your Mac. Your audio is never uploaded, and there is no telemetry.
Open
MIT licensed. Read the code, build it yourself, or help improve it.
Side by side
How Hush compares
Feature by feature, with the gaps left in.
- ✓ Yes
- ◐ Partly
- ✗ No
- ? Not stated on their site
- Hover or tap a cell for details
| Feature | Hush | Wispr Flow | Superwhisper | VoiceInk | OpenWhispr |
|---|---|---|---|---|---|
| Price & openness | |||||
| PricePaid plan, monthly unless noted | Free | $15/moPro plan; $12/mo billed yearly. | $8.49/moOr $84.99/yr, or $249.99 lifetime. | $25 one-timeFor 1 Mac; $39 for 2 Macs, $49 for 3. | $8/moPro plan, or $80/yr. Only needed for their hosted cloud. |
| Free versionWhat you get without paying | Everything | 2,000 words a weekOn desktop; 1,000 words a week on iPhone. | Local modelsUnlimited dictation with local models, no AI clean-up. | 7-day trialFree if you build it yourself from the GPL source. | Local modelsUnlimited with local models or your own keys; 2,000 cloud words a week. |
| Open sourceLicense, if the code is public | MIT | ✗Terms forbid decompiling to access source code. | ✗Terms forbid extracting the app's source code. | GPL-3.0 | MIT |
| AccountDo you have to sign in? | Not needed | Required | Not neededActivates with a license key. | Not neededActivates with a license key. | ◐Only for their hosted cloud; local models and your own keys need no account. |
| Privacy | |||||
| On-device speech recognitionYour voice is turned into text on your computer | ✓Parakeet, Whisper, Moonshine, Canary and other open models run locally. | ✗Cloud only. | ✓Cloud models also available. | ✓Cloud providers optional. | ✓Cloud models also available. |
| Works offline | ✓Only the optional AI clean-up with a cloud provider needs a connection. | ✗Their docs say an internet connection is required. | ✓ | ✓With local models. | ✓With local models. |
| Audio sent to a serverWhen your recording leaves your computer | Never | Always | By defaultCloud models are the default on Pro; local models keep audio on your Mac. | Only with cloud models | Only with cloud models |
| Usage analyticsTelemetry about how you use the app | None | YesUsage data is collected even with “Never store my data” on. | YesThe App Store privacy label lists diagnostics and a device ID; their privacy policy says no usage data is collected. | ?Their privacy policy doesn't say either way. | Opt-inOff by default. |
| Dictation | |||||
| Push to talkHold a key, speak, release | ✓ | ✓ | ✓ | ✓ | ✓ |
| Hands-free modeTap once to start, again to stop | ✓A quick tap latches; holding works as push to talk. | ✓ | ✓ | ✓ | ✓ |
| Live previewSee words appear while you speak | ✓ | ✗Text is inserted after you stop speaking. | ✓Pro, with selected models. | ◐Only with streaming models, not local Whisper. | ◐Only with streaming models. |
| Works in any app | ✓ | ✓ | ✓ | ✓ | ✓ |
| Languages | 9999 with Whisper models; 25 European languages with the default Parakeet V3. | 100+ | 100+Depends on the model. | ?Multilingual models, but no total is stated. | 60Selectable in the app; their website says 100+. |
| AI clean-up | |||||
| AI clean-upFixes punctuation, filler words and formatting | ✓Optional, with your own provider key or Apple Intelligence. | ✓Built in, through their cloud. | ✓Pro feature. | ✓ | ✓ |
| Bring your own keyUse your own AI provider account | ✓Groq, OpenAI, Anthropic, OpenRouter, Cerebras or any OpenAI-compatible endpoint. | ? | ✓Pro feature. | ✓ | ✓ |
| On-device AI optionClean-up with a local model or Apple Intelligence | ✓Apple Intelligence on supported Macs. | ? | ✓Their S1-mini model: English, Message and Mail modes. | ✓VoiceInk Refine or Ollama. | ✓Bundled local models, Ollama or LM Studio. |
| Handles self-corrections“At 4, no, 5” becomes “At 5” | ✓Part of the AI clean-up. | ✓ | ? | ✓ | ✓ |
| Per-app stylesWrites differently in chat, email or code | ✓Default, Message, Email, Code and Raw. | ✓By app category, English only. | ✓ | ✓ | ? |
| Custom vocabularyNames and terms spelled your way | ✓ | ✓ | ✓ | ✓ | ✓ |
| Voice snippetsA short phrase expands into longer text | ✓ | ✓ | ✓ | ✓Through word replacements. | ✓ |
| Power features | |||||
| Edit selected text by voice“Make this more formal” | Not yet, Coming next | ✓Command Mode, Pro plan. | ✓ | ✓ | ✓Through the voice agent shortcut. |
| Context awarenessUses text on screen or around the cursor | Not yet, Coming next | ✓ | ✓Pro feature. | ✓Selected text, clipboard and screen text. | ✓In the voice agent, not plain dictation. |
| Learns from your corrections | Not yet, Coming next | ✓ | ? | ✓ | ✓ |
| Pauses media while dictating | ✓Pause, lower the volume or mute; resumes only what it paused. | ✓Mutes music on desktop. | ✓ | ✓ | ✓ |
| File transcriptionTurn audio or video files into text | Not yet, Coming later | ✗ | ✓Pro feature. | ✓ | ✓ |
| History | ✓Stored only on your Mac. | ✓ | ✓ | ✓ | ✓ |
| Platform | |||||
| Mac (Apple Silicon) | ✓ | ✓ | ✓ | ✓ | ✓ |
| Mac (Intel) | Not yet, Coming later | ✓ | ✓Works best with cloud models on Intel. | ✗ | ✓ |
| Windows | Not yet, Coming later | ✓ | ✓ | ✗ | ✓ |
| iPhone | ✗Not on the roadmap yet. | ✓ | ✓ | ✓ | ✓ |
Show more apps: Spokenly, Willow, Apple DictationHide Spokenly, Willow and Apple Dictation
| Feature | Hush | Spokenly | Willow | Apple Dictation |
|---|---|---|---|---|
| Price & openness | ||||
| PricePaid plan, monthly unless noted | Free | $9.99/moPro plan, or $99.99/yr; a one-time Local+ plan is $29. | $15/moPro plan; $12/mo billed yearly. | FreeBuilt into macOS. |
| Free versionWhat you get without paying | Everything | Local modelsUnlimited with local models or your own API keys. | LimitedUnlimited dictation on a lighter model, plus 20 uses a week of their Scribe feature. | Everything |
| Open sourceLicense, if the code is public | MIT | ✗Terms say the app is owned by Spokenly and protected by copyright. | ✗Terms prohibit reverse engineering. | ✗ |
| AccountDo you have to sign in? | Not needed | Not neededSign-in only to activate a purchase made on their website. | Required | ? |
| Privacy | ||||
| On-device speech recognitionYour voice is turned into text on your computer | ✓Parakeet, Whisper, Moonshine, Canary and other open models run locally. | ✓Cloud models are the default; local models optional. | ◐Cloud by default; local recognition only in Offline mode (Pro). | ◐On Apple Silicon for 48 languages; otherwise Apple's servers. |
| Works offline | ✓Only the optional AI clean-up with a cloud provider needs a connection. | ✓With local models. | ◐Offline mode is a Pro feature. | ◐Only for on-device languages on Apple Silicon. |
| Audio sent to a serverWhen your recording leaves your computer | Never | By defaultCloud models are the default; local models keep audio on your device. | By defaultExcept in Offline mode (Pro). | SometimesOn Intel Macs, for languages without on-device support, and in search boxes. |
| Usage analyticsTelemetry about how you use the app | None | On by defaultTurned off only by Local Only Mode. | Yes | Opt-inStoring audio is opt-in; server requests still log device and performance data. |
| Dictation | ||||
| Push to talkHold a key, speak, release | ✓ | ✓ | ✓ | ?Apple documents press-to-start only; holding the Microphone key opens Siri. |
| Hands-free modeTap once to start, again to stop | ✓A quick tap latches; holding works as push to talk. | ✓ | ✓ | ✓ |
| Live previewSee words appear while you speak | ✓ | ✓With streaming models. | ? | ? |
| Works in any app | ✓ | ✓ | ✓ | ✓ |
| Languages | 9999 with Whisper models; 25 European languages with the default Parakeet V3. | 100+ | 100+ | 6348 of them on-device. |
| AI clean-up | ||||
| AI clean-upFixes punctuation, filler words and formatting | ✓Optional, with your own provider key or Apple Intelligence. | ✓ | ✓ | ?Only automatic punctuation is documented for Dictation. |
| Bring your own keyUse your own AI provider account | ✓Groq, OpenAI, Anthropic, OpenRouter, Cerebras or any OpenAI-compatible endpoint. | ✓ | ? | ? |
| On-device AI optionClean-up with a local model or Apple Intelligence | ✓Apple Intelligence on supported Macs. | ✓Needs Apple Silicon and macOS 26. | ? | ? |
| Handles self-corrections“At 4, no, 5” becomes “At 5” | ✓Part of the AI clean-up. | ✓ | ✓ | ? |
| Per-app stylesWrites differently in chat, email or code | ✓Default, Message, Email, Code and Raw. | ✓ | ✓ | ? |
| Custom vocabularyNames and terms spelled your way | ✓ | ✓ | ✓ | ? |
| Voice snippetsA short phrase expands into longer text | ✓ | ◐No snippet feature, but word replacements can expand text. | ✓ | ? |
| Power features | ||||
| Edit selected text by voice“Make this more formal” | Not yet, Coming next | ? | ✓ | ✗Dictation commands cover punctuation and formatting; Apple points to Voice Control. |
| Context awarenessUses text on screen or around the cursor | Not yet, Coming next | ✓ | ✓ | ? |
| Learns from your corrections | Not yet, Coming next | ? | ✓ | ? |
| Pauses media while dictating | ✓Pause, lower the volume or mute; resumes only what it paused. | ✓ | ✓Mutes background audio rather than pausing it. | ? |
| File transcriptionTurn audio or video files into text | Not yet, Coming later | ✓ | ? | ? |
| History | ✓Stored only on your Mac. | ✓ | ✓ | ? |
| Platform | ||||
| Mac (Apple Silicon) | ✓ | ✓ | ✓ | ✓ |
| Mac (Intel) | Not yet, Coming later | ✓ | ? | ✓Recognition runs on Apple's servers on Intel Macs. |
| Windows | Not yet, Coming later | ✓ | ✓ | ✗ |
| iPhone | ✗Not on the roadmap yet. | ✓ | ✓ | ✓ |
Price & openness
HushWispr Flow- Price
- Free
- $15/moPro plan; $12/mo billed yearly.
- Free version
- Everything
- 2,000 words a weekOn desktop; 1,000 words a week on iPhone.
- Open source
- MIT
- ✗Terms forbid decompiling to access source code.
- Account
- Not needed
- Required
Privacy
HushWispr Flow- On-device speech recognition
- ✓Parakeet, Whisper, Moonshine, Canary and other open models run locally.
- ✗Cloud only.
- Works offline
- ✓Only the optional AI clean-up with a cloud provider needs a connection.
- ✗Their docs say an internet connection is required.
- Audio sent to a server
- Never
- Always
- Usage analytics
- None
- YesUsage data is collected even with “Never store my data” on.
Dictation
HushWispr Flow- Push to talk
- ✓
- ✓
- Hands-free mode
- ✓A quick tap latches; holding works as push to talk.
- ✓
- Live preview
- ✓
- ✗Text is inserted after you stop speaking.
- Works in any app
- ✓
- ✓
- Languages
- 9999 with Whisper models; 25 European languages with the default Parakeet V3.
- 100+
AI clean-up
HushWispr Flow- AI clean-up
- ✓Optional, with your own provider key or Apple Intelligence.
- ✓Built in, through their cloud.
- Bring your own key
- ✓Groq, OpenAI, Anthropic, OpenRouter, Cerebras or any OpenAI-compatible endpoint.
- ?
- On-device AI option
- ✓Apple Intelligence on supported Macs.
- ?
- Handles self-corrections
- ✓Part of the AI clean-up.
- ✓
- Per-app styles
- ✓Default, Message, Email, Code and Raw.
- ✓By app category, English only.
- Custom vocabulary
- ✓
- ✓
- Voice snippets
- ✓
- ✓
Power features
HushWispr Flow- Edit selected text by voice
- Not yet, Coming next
- ✓Command Mode, Pro plan.
- Context awareness
- Not yet, Coming next
- ✓
- Learns from your corrections
- Not yet, Coming next
- ✓
- Pauses media while dictating
- ✓Pause, lower the volume or mute; resumes only what it paused.
- ✓Mutes music on desktop.
- File transcription
- Not yet, Coming later
- ✗
- History
- ✓Stored only on your Mac.
- ✓
Platform
HushWispr Flow- Mac (Apple Silicon)
- ✓
- ✓
- Mac (Intel)
- Not yet, Coming later
- ✓
- Windows
- Not yet, Coming later
- ✓
- iPhone
- ✗Not on the roadmap yet.
- ✓
Based on each product's website, October 2026. Spotted something wrong? Open an issue.
Switching
Why people switch to Hush
- 1
The subscription adds up
Paid dictation apps cost up to $180 a year. Hush has no paid tier, so there is nothing to upgrade to and no word limit to run into.
- 2
You'd rather your voice stayed home
Hush recognizes speech on your Mac and works offline. AI clean-up is optional, and when it's on, only the text goes to the provider you picked.
- 3
You want to see how it works
Every line that touches your voice is on GitHub. To be fair, some paid apps do more today, such as iPhone apps and editing text by voice. Editing by voice is on our roadmap.
See what's coming on the public roadmap.
Alternatives
Switching from one app in particular?
Each page goes deeper on one app: why people leave it, where it's still ahead, and how to switch.
Wispr Flow
The word limit, then the bill.
Hush as a Wispr Flow alternativeSuperwhisper
AI clean-up is a Pro feature.
Hush as a Superwhisper alternativeVoiceInk
A license for each Mac.
Hush as a VoiceInk alternative
FAQ
Good questions
How is Hush different from Wispr Flow?
Hush is free and open source, and recognizes speech on your Mac. Wispr Flow is a paid monthly service that processes speech in the cloud, and it is available on more platforms. See Hush as a Wispr Flow alternative or the full comparison.
How is Hush different from Superwhisper?
Both can recognize speech on your Mac. Hush is free and open source, and its AI clean-up and per-app styles cost nothing beyond your own provider key, or nothing at all with Apple Intelligence. Superwhisper is a paid app whose free version covers local dictation without AI clean-up, and it does more today, such as transcribing files and running on Windows and iPhone. See Hush as a Superwhisper alternative.
So what's the catch?
There isn't one. Hush runs on your own Mac, so there are no servers to pay for and nothing to charge you for. There are no ads and no data to sell, because Hush never sees your data. The code is public, so you can check.
Does it work offline?
Yes. Once you have downloaded a speech model, dictation works without an internet connection. Only AI clean-up with a cloud provider needs a connection.
You talk faster than you type.
About three times faster. Hush turns that into text you can actually send, for free, without your voice leaving your Mac.
Apple Silicon Macs · Free and open source