Wispr Flow vs Superwhisper: AI Dictation Apps in 2026

The first thing dictation broke for me was copy and paste.

I had copied a block of JSON out of a log, clicked into a chat with a coding agent, and dictated "here is the error, tell me which field is null." Then I pressed Cmd+V to paste the JSON underneath. What came out was my own sentence, a second time. The dictation app had put its transcript on the clipboard to insert it, and my JSON was gone.

Both of the apps in this comparison insert text that way. Every review I read before writing this one compares accuracy, price and platforms, which matter, and then says nothing about the clipboard, which is the part you will actually trip over in week one. So this page covers both.

If you want to turn recordings and meetings into text after the fact, that is a different job, and the transcription tools roundup is where to start. Our Buzz review covers the free, fully offline Whisper option for files. This page is about live dictation: you hold a key, you talk, words appear wherever your cursor is.

What each one actually is

Wispr Flow is a cloud dictation service with very good manners. You trigger it, speak, and a server transcribes your audio and runs it through a cleanup pass that adds punctuation, drops the "um"s and the false starts, and formats lists when you obviously meant a list. The company's own data controls page is blunt about the architecture: transcription always happens in the cloud. There is no offline mode at any price, including Enterprise.

Superwhisper is a toolkit. It splits dictation into two stages, voice to text and then optional language-model post-processing, and lets you decide separately where each one runs. Pick a local Whisper or Parakeet model and your audio never leaves the machine. Pick a cloud model like Nova 3 and it does. Then you choose whether a language model rewrites the result, and which one, through "modes" you can customise down to the system prompt.

The short answer

	Wispr Flow	Superwhisper
Where transcription runs	Cloud only	Your choice: on-device or cloud
Works offline	No	Yes, with a local model
Platforms	Mac, Windows, iPhone, Android	Mac, Windows, iOS, Android
Free tier	2,000 words a week on desktop, 1,000 on iPhone	Permanent, all Whisper models included
Paid	$15 a month, or $12 a month billed yearly	$8.49 a month, $84.99 a year, or $249.99 lifetime
AI cleanup	On by default, tuned, little to configure	Per mode, fully configurable, more setup
Default clipboard behaviour	Restores your old clipboard about half a second after pasting	Overwrites it, unless you switch on "Restore clipboard"

Flow is better out of the box. Superwhisper is better once you have spent an evening with it.

Wispr Flow, the one that needs no setup

The cleanup pass is the product. I can ramble a paragraph, change my mind halfway through a sentence ("send it Tuesday, no, Wednesday"), and Flow usually writes the Wednesday version without the correction left in. That is a hard thing to do well and Flow does it better than anything I have configured myself. For Slack replies and email it is almost unfair.

It also has the widest reach. Mac, Windows, iPhone and Android all share one account and one personal dictionary, so a client's odd product name learned on the laptop is spelled right on the phone.

The costs are all consequences of the cloud. On a train with patchy signal it stalls. Every word you say goes to a server, and if you switch on Context Awareness, so does relevant text from the app window you are in, which is how it gets names and variable names right. For a lot of people that trade is fine. For anyone dictating patient notes or a legal draft, it ends the conversation before it starts, though Flow's Enterprise tier does advertise SOC 2 Type II and a HIPAA BAA.

Terminals are the other rough edge. Flow's own help docs warn that its automatic capitalisation and punctuation can mangle shell commands, and that WSL windows and remote SSH sessions need a "Paste last transcript" workaround (Ctrl+Cmd+V on a Mac) instead of a direct paste. I stopped dictating commands entirely. I still dictate everything around them.

Superwhisper, the one you tune

Superwhisper asks you to make decisions Flow makes for you, and the first day is worse for it.

The free plan includes every Whisper model, up to Large v3 Turbo, running locally. That alone is a better free offer than Flow's weekly word cap if you have Apple Silicon or a decent GPU on Windows. Pro adds Parakeet, Cohere Transcribe (which Superwhisper's own docs recommend as the most accurate local model for sensitive work), and the cloud models.

The modes are where it pulls ahead for anyone with a specialist vocabulary. I have one mode that transcribes with no language model at all, for anything I would not want rewritten. Another takes my rambling and turns it into a tight prompt for a coding agent, with a system prompt that tells it never to "fix" file paths, flags or function names. Flow's cleanup cannot be told that. It has opinions about capital letters.

Superwhisper also documents a coding-agents setup and ships a CLI and an MCP server, which says plenty about who it thinks its users are.

The downside is that the defaults are plain. Out of the box the cleanup is weaker than Flow's, and getting it as good takes prompt-writing that most people will never do.

What both apps do with your clipboard

Both apps put your words where the cursor is the same way: they write the transcript to the system clipboard, then send a paste. It is the only insertion method that works across nearly every app. It also means a dictation app is, mechanically, something that writes to your clipboard dozens of times a day.

It is an old trick, and it holds up remarkably well until the day it doesn't.

What you copied can be overwritten. Flow handles this well by default. Its help centre says it restores your previous clipboard about half a second after pasting, and keeps anything new you copy in that window. Superwhisper's advanced settings state the opposite default plainly: "your result replaces whatever was on your clipboard." Set Clipboard behaviour to Restore clipboard on day one. That one setting would have saved my JSON.

What you copied can be read. Superwhisper's context awareness can include clipboard content copied within three seconds before you start recording, or during the dictation, along with selected text and the active window's fields, and hand it to that mode's language model. Most built-in modes leave it off. If you turn it on and use a cloud model, the API key you just copied from a dashboard rides along with your sentence. Flow's Context Awareness reads text from the active app instead. Neither is a scandal. Both are worth knowing before you dictate next to a password manager.

Other apps notice. Coding agents are where dictation is most useful and where this shows. An open issue on OpenAI's Codex repo (#11103) describes Superwhisper's clipboard auto-paste going wrong in the Codex terminal UI after several prompts, with the suggested workaround being Superwhisper's "Simulate keypresses" delivery, which types the text instead of pasting it. Another, on the Claude Code repo (#90747), reports the Windows desktop app showing a clipboard warning on every single Wispr Flow dictation, because it cannot tell a dictation tool from something malicious. Both are open as I write this.

None of the top-ranking comparisons mention any of this. You will meet it within a week.

Which one to install

Install Wispr Flow if you write mostly prose (email, Slack, docs) across several devices, you are always online, and you want it to be good on the first afternoon. Pay the $12 a month yearly rate if you stick with it past the free week.

Install Superwhisper if any of your words should not leave the machine, if you dictate into terminals and coding agents, or if you are on a Mac with enough memory to run a large local model. Turn on Restore clipboard before anything else. The lifetime licence makes sense for anyone who knows within a month that they are keeping it.

If you dictate prompts to AI models all day, as I now do, Superwhisper. The ability to tell the cleanup model what not to touch matters more to me than Flow's better default polish. That is the same reason I keep a local model running through Ollama: some of what I say to a computer is nobody else's business.

Frequently Asked Questions

Is Wispr Flow or Superwhisper better?

Wispr Flow is better out of the box, with stronger automatic cleanup and apps on Mac, Windows, iPhone and Android. Superwhisper is better for privacy and control, because it can transcribe entirely on-device with local Whisper or Parakeet models and lets you configure the AI cleanup per mode. Flow is cloud-only and does not work offline.

Does Wispr Flow work offline?

No. Wispr Flow's own data controls page states that transcription always occurs in the cloud. Superwhisper works offline when you select a local voice model.

How much do Wispr Flow and Superwhisper cost?

Wispr Flow has a free tier of 2,000 words a week on desktop, and Pro costs $15 a month or $12 a month billed yearly. Superwhisper has a permanent free tier that includes all Whisper models, and Pro costs $8.49 a month, $84.99 a year, or $249.99 for a lifetime licence.

Why does my clipboard change after I dictate?

Both apps insert text by placing the transcript on the clipboard and pasting it. Wispr Flow restores your previous clipboard about half a second later by default. Superwhisper replaces your clipboard by default, which you can change by setting Clipboard behaviour to Restore clipboard in its advanced settings.