Compare
YapScribe and Wispr Flow, row by row.
Two screen-aware dictation apps that answer the same question differently. This page is the difference, with a source against every claim we make about theirs.
What each one does
| What | YapScribe | Wispr Flow |
|---|---|---|
| What it costs | One payment. A licence covers two Macs and a year of updates; after that the app keeps working and newer builds need an optional update pass. The rung that is selling right now is on our pricing page. | A free tier, then $15 per user per month, or $12 per user per month on an annual plan. Growth is $23, or $18 annual. Enterprise is a custom annual contract. Source1 |
| Screen context, and where the screenshot goes | Your screen is read on your Mac. The screenshot is held in memory for one dictation and then released — never written to disk, and never put into any request, on any setting. A short list of words read from it travels only when you have cloud polishing on. | Their Context Awareness page lists what context holds, including “a screenshot”, and says context is collected locally and sent to Wispr with each dictation request unless Privacy Mode is on. Their Data Controls page describes Context Awareness as using “relevant text data from the active app window”. Context Awareness is listed as full on Mac, partial on Windows and limited to the focused text field on iOS. |
| Speech to text | On your Mac — by a Whisper model you download once, or, on macOS 26 and later, by Apple’s speech model that ships inside macOS. The audio is never uploaded, on either plan. | Their privacy page states that “transcription always happens in the cloud”. Source2 |
| Bringing your own API key | OpenAI, Anthropic, Groq, Google Gemini or xAI with a key of your own, or any endpoint that speaks OpenAI’s chat format — including Ollama or LM Studio on your own Mac. You pay that company directly and us nothing further for polishing. | Not stated publicly. Nothing on their pricing page, their home page or their “What is Wispr Flow?” article mentions supplying your own provider key. |
| Using it offline | Speech to text and your wordlist always work offline — both run on your Mac. Polishing works offline too with it switched off, or pointed at a model running on your own machine. | Their privacy page states that transcription always happens in the cloud, and their “What is Wispr Flow?” article states that Flow requires an internet connection for voice transcription. Their Android requirements page states that offline transcription is not available. |
| Polishing without sending anything | Three ways, on every plan: polishing off, a model you run yourself on the same Mac, or — on macOS 26 and later, with Apple Intelligence switched on — the model built into macOS, which makes no network request for a dictation at all. That last one is the least accurate polish YapScribe offers: it does not fix mid-sentence corrections, which the cloud models do. | Not stated publicly. Nothing on their privacy page, their home page or their “What is Wispr Flow?” article describes formatting or polishing running on your own machine, or an option to keep it there. Their privacy page states that transcription always happens in the cloud. |
| Languages | Whichever languages the Whisper model you downloaded handles. The app detects which one you are speaking rather than asking. Apple’s speech model, the optional one on macOS 26 and later, listens in your Mac’s own language instead of detecting. We do not publish a number. | Stated as 100+ languages. Their home page says Flow detects the language you are speaking automatically, and its FAQ recommends selecting the language you are speaking rather than relying on auto-detect. Their “What is Wispr Flow?” article says to choose your languages in each platform’s settings. |
| Fixing yourself mid-sentence | Saying “scratch that” or “forget that” inside a sentence discards what came before it, and false starts are cleaned up with the fillers. Saying only “scratch that” — the whole utterance and nothing else — within a minute and still in the same field of the same app takes the previous dictation back out with one undo. Outside those guards the words are typed as you said them. | Backtrack is described as removing filler words, false starts and self-corrections, triggered by phrases such as “actually” or “scratch that”, or by restating what you meant. Source8 |
| A dictionary that learns | Learns names from what is on your screen and from the edits you make to a result, on your Mac, and you review what it learned in Settings. It does not sync between your Macs. In a cloud-polish set-up the list travels with the correction request, as the Trust ledger says. | Described as learning your names and jargon as you go, with a dictionary you can also add to yourself. Dictionary entries are stated to sync across your devices. |
| A different tone per app | A built-in tone per kind of app: a chat message is not an email and neither is a document. In Xcode and the Terminal it does not rewrite you at all. On the Apps tab in Settings you can change which kind one app counts as, turn polishing off for it, or write your own instructions for it, which go with every polished dictation into that app. | Described as adapting to how you write in different apps, with a style you can set for messages, work chats, emails and more. Source3 |
| Whispering | Quiet speech has its levels fixed on your Mac before the model hears it, with nothing to switch on. Normal speech is left at the level you spoke it. | Described as handling speech as quiet as a whisper. Source3 |
| A history of what you dictated | On your Mac: the finished text and the raw transcript behind it, with a “show what I actually said” view and a clear-history control. No audio is kept, nothing syncs, and our servers never hold any of it. | Transcripts are stated to be stored locally on each device. A separate Dictation Cloud Storage setting is described as storing transcripts, audio and dictation history on Wispr’s servers. |
| Where it runs | Apple Silicon Macs — M1 and newer — on macOS 14 Sonoma or later. There is no Intel build, and there is no Windows, iPhone or Android app. | Mac on macOS 12 or later, Windows 10 and 11, iPhone on iOS 18.3 or later, and Android 13 to 16. |
What it costs
- YapScribe
- One payment. A licence covers two Macs and a year of updates; after that the app keeps working and newer builds need an optional update pass. The rung that is selling right now is on our pricing page.
- Wispr Flow
A free tier, then $15 per user per month, or $12 per user per month on an annual plan. Growth is $23, or $18 annual. Enterprise is a custom annual contract.
Source1
Screen context, and where the screenshot goes
- YapScribe
- Your screen is read on your Mac. The screenshot is held in memory for one dictation and then released — never written to disk, and never put into any request, on any setting. A short list of words read from it travels only when you have cloud polishing on.
- Wispr Flow
Their Context Awareness page lists what context holds, including “a screenshot”, and says context is collected locally and sent to Wispr with each dictation request unless Privacy Mode is on. Their Data Controls page describes Context Awareness as using “relevant text data from the active app window”. Context Awareness is listed as full on Mac, partial on Windows and limited to the focused text field on iOS.
Speech to text
- YapScribe
- On your Mac — by a Whisper model you download once, or, on macOS 26 and later, by Apple’s speech model that ships inside macOS. The audio is never uploaded, on either plan.
- Wispr Flow
Their privacy page states that “transcription always happens in the cloud”.
Source2
Bringing your own API key
- YapScribe
- OpenAI, Anthropic, Groq, Google Gemini or xAI with a key of your own, or any endpoint that speaks OpenAI’s chat format — including Ollama or LM Studio on your own Mac. You pay that company directly and us nothing further for polishing.
Using it offline
- YapScribe
- Speech to text and your wordlist always work offline — both run on your Mac. Polishing works offline too with it switched off, or pointed at a model running on your own machine.
Polishing without sending anything
- YapScribe
- Three ways, on every plan: polishing off, a model you run yourself on the same Mac, or — on macOS 26 and later, with Apple Intelligence switched on — the model built into macOS, which makes no network request for a dictation at all. That last one is the least accurate polish YapScribe offers: it does not fix mid-sentence corrections, which the cloud models do.
Languages
- YapScribe
- Whichever languages the Whisper model you downloaded handles. The app detects which one you are speaking rather than asking. Apple’s speech model, the optional one on macOS 26 and later, listens in your Mac’s own language instead of detecting. We do not publish a number.
- Wispr Flow
Stated as 100+ languages. Their home page says Flow detects the language you are speaking automatically, and its FAQ recommends selecting the language you are speaking rather than relying on auto-detect. Their “What is Wispr Flow?” article says to choose your languages in each platform’s settings.
Fixing yourself mid-sentence
- YapScribe
- Saying “scratch that” or “forget that” inside a sentence discards what came before it, and false starts are cleaned up with the fillers. Saying only “scratch that” — the whole utterance and nothing else — within a minute and still in the same field of the same app takes the previous dictation back out with one undo. Outside those guards the words are typed as you said them.
- Wispr Flow
Backtrack is described as removing filler words, false starts and self-corrections, triggered by phrases such as “actually” or “scratch that”, or by restating what you meant.
Source8
A dictionary that learns
- YapScribe
- Learns names from what is on your screen and from the edits you make to a result, on your Mac, and you review what it learned in Settings. It does not sync between your Macs. In a cloud-polish set-up the list travels with the correction request, as the Trust ledger says.
A different tone per app
- YapScribe
- A built-in tone per kind of app: a chat message is not an email and neither is a document. In Xcode and the Terminal it does not rewrite you at all. On the Apps tab in Settings you can change which kind one app counts as, turn polishing off for it, or write your own instructions for it, which go with every polished dictation into that app.
- Wispr Flow
Described as adapting to how you write in different apps, with a style you can set for messages, work chats, emails and more.
Source3
Whispering
- YapScribe
- Quiet speech has its levels fixed on your Mac before the model hears it, with nothing to switch on. Normal speech is left at the level you spoke it.
- Wispr Flow
Described as handling speech as quiet as a whisper.
Source3
A history of what you dictated
- YapScribe
- On your Mac: the finished text and the raw transcript behind it, with a “show what I actually said” view and a clear-history control. No audio is kept, nothing syncs, and our servers never hold any of it.
Where it runs
- YapScribe
- Apple Silicon Macs — M1 and newer — on macOS 14 Sonoma or later. There is no Intel build, and there is no Windows, iPhone or Android app.
What we do not have yet
A comparison page that only lists wins is an advert. These are the reasons to pick the other one.
- A Windows, iPhone or Android app. YapScribe is Apple Silicon Macs on macOS 14 or later, and nothing else.
- Anything for a team: no single sign-on, no central billing, no admin console. We sell one licence to one person.
- A meeting notetaker. YapScribe dictates into the field you have open; it does not sit in your calls.
- Text that appears as you speak. Yours arrives when you stop, as one paste you can undo in one step.
- Your wordlist and learned words syncing between your own Macs. They live on each Mac today.
- Voice commands beyond fixing yourself and taking back the last dictation. “Make that shorter” is not in this build.
Sources
Wispr Flow’s own public pages, in the order they are cited above, each with the date it was read.
- Wispr Flow — Pricing
- Wispr Flow — Privacy
- Wispr Flow — home page
- Wispr Flow Help Center — Context Awareness
- Wispr Flow — Data Controls
- Wispr Flow Help Center — What is Wispr Flow?
- Wispr Flow Help Center — Supported devices and system requirements
- Wispr Flow Help Center — Smart Formatting and Backtrack
- Wispr Flow Help Center — Teach Flow your words with the dictionary
- Wispr Flow Help Center — Delete transcripts and history
- Wispr Flow Help Center — Manage data sharing, cloud storage and local history
- Wispr Flow Help Center — Android system requirements
YapScribe’s own column is backed by this repository rather than by a link: what the app sends is the app’s own privacy copy, and the table on the home page is the same claims in full — the Trust ledger.
Try it for seven days before deciding any of this.
Download for macOS