Two true sentences that sound like they disagree
Voice dictation really is about three times faster than typing for a lot of people. And switching to it really will make your working day about one point two times faster. Both of those are correct, they are computed from the same numbers, and the gap between them is the single most useful thing on this page.
The reason is not subtle once you see it. Speeding up part of a job speeds up the job by the share that part took. If putting words down is a quarter of your day, then a quarter is the absolute most that any typing tool could ever give you back, even one that worked instantly and perfectly. Every claim you read is about the first number. Everything you feel is the second.
The formula
time saved = writing time × (1 − effective typing ÷ effective dictation)
share of day saved = time saved ÷ hours worked
Cleanup divides rather than subtracts, because it is extra time spent on top of getting the words down. At 25%, five minutes of dictation costs you six and a quarter, so the rate you actually achieve is the raw one divided by 1.25. Both sides of the comparison get the same treatment, for reasons worth a section of their own.
Which cleanup is which, and why there are two boxes
They are separate jobs, and mixing them up is easy. Cleaning up after dictation is fixing what the recognition got wrong: names, punctuation, technical terms, and the sentences that came out as you said them rather than as you meant them. Cleaning up after typing is fixing your own typos, and specifically the ones you did not catch while you were typing.
That second box defaults to zero, and the reason is worth knowing rather than guessing at. Published typing speeds come from typing tests, where you backspace and fix as you go, and all that fixing happens inside the measured minute. So a 40 word a minute figure is already net of the mistakes you catch in flight. What it does not include is the typos you sail straight past and clean up later in a proofreading pass. If that describes you, the box is there to say so, and it is doing real work: it is the difference between the speed you type at and the speed you finish text at.
There is a second reason not to assume typing is the tidy one. Dictation tools that run the transcript through a language model can hand back text needing almost no correction, and if you use one of those while also being a heavy backspacer, the side with the bigger cleanup bill is the keyboard. Nothing about the arithmetic minds which way round that goes, but almost every published comparison assumes the old way, because it was written when dictation meant raw transcription.
The two boxes are also where a lot of the answer lives, which is easy to miss because the raw speeds are the numbers everybody argues about. Take the worked example below and change nothing except the cleanup: 25% after dictation and none after typing gives 3.00 times. Swap it to 5% after dictation and 15% after typing, which is a fair description of a modern tool and a fast messy typist, and the same 40 and 150 words a minute become 4.11 times. That is a 37% swing with both speeds untouched.
Worked example
You type 40 words a minute, speak at 150, spend 25% of the dictation time tidying up and nothing cleaning up your typing, and 25% of an eight hour day putting words down.
Effective dictation speed: 150 ÷ 1.25 = 120 words a minute, which is 3.00 times your typing.
Two hours of typing produces 4,800 words. Dictating the same 4,800 words takes 40 minutes.
You save 1h 20m a day, 6h 40m a week, and 320 hours a year, which is about eight working weeks.
And yet your day is only 1.20 times faster, because you saved 80 minutes out of 480. The tool tripled. The day went up a fifth.
Where the famous "3x faster" number actually comes from
Nearly every article about dictation cites a Stanford study finding speech about three times faster than typing, with a 20.4% lower error rate in English. That study is real and the finding is solid. It is also, specifically, about typing on a phone.
Thumb-typing on a touchscreen runs somewhere around 30 to 38 words a minute. Speech at 150 is roughly four times that before any editing, so a threefold result after real-world messiness is exactly what you would expect. The number is not inflated. It is simply answering the question "is talking faster than thumbs", and then getting quoted at people sitting at desks with full-size keyboards.
Run the same comparison against a good touch typist and it looks very different:
| How you type | Typing speed | Effective dictation at 150 wpm | Faster by |
|---|---|---|---|
| Thumbs on a phone | 34 wpm | 120 wpm | 3.5x |
| Average adult | 40 wpm | 120 wpm | 3.0x |
| Competent office typist | 60 wpm | 120 wpm | 2.0x |
| Good touch typist | 75 wpm | 120 wpm | 1.6x |
| Very fast typist | 120 wpm | 120 wpm | no gain |
So the group these tools help least is fast touch typists, which is close to the opposite of what the advertising implies. If you are quick on a keyboard, the honest expectation is a modest gain rather than a transformation, and this calculator will tell you plainly when dictation would actually be slower for you.
The subscription argument is better than the marketing one
Vendors like to quote the big annual figure. On the worked example that is 320 hours, and at $50 an hour it is $16,000 a year against maybe $180 for the tool. Those numbers are arithmetically fine and they are also the least convincing way to make the case, because saved hours are not money that appears in your account. An hour you get back is worth an hour's pay only if something valuable fills it.
The stronger argument is the one nobody leads with. Turn it around and ask how little the tool has to save before it pays for itself: on those same numbers, about five minutes a week. That is a bar almost any working writer clears in a single afternoon, and it does not require you to believe any claim about tripling your output. If a tool is cheap relative to an hour of your time, the decision is mostly about whether you enjoy using it.
The dictation tools people are actually using in 2026
This is a list, not a ranking, and it is grouped by where your voice gets processed rather than by which is best, because that is the choice that actually decides which one suits you. We are not paid by anything on this list and none of these are affiliate links. Prices are what the vendor published in August 2026.
Cloud tools that clean up as they go
These record, transcribe, then pass the raw text through a language model that strips filler words, adds punctuation and adapts to what you are writing into. That last step is the reason the cleanup box above can sit near zero for some people, and it is a genuinely new thing rather than marketing.
- Wispr Flow: works system wide across Mac, Windows, iPhone and Android, in over 100 languages. Free up to 2,000 words a week on desktop and 1,000 on iPhone, with unlimited on Android. Pro is $15 a month, or $12 if you pay for the year.
- Aqua Voice: Mac, Windows and iOS, aimed at people dictating technical prose and code. 1,000 free words to try, then $8 a month, or $24 for a tier with a real time mode.
Tools that never send your voice anywhere
These run the speech model on your own machine. Worth the trade if you dictate anything you would not paste into a web form: client notes, patient details, unreleased work, other people's private business.
- Superwhisper: Mac, Windows and iOS, and it works with no internet connection at all. There is a free tier with unlimited use of the local models, and Pro is $8.49 a month with yearly and lifetime options.
- MacWhisper: Mac, iPhone and iPad, transcription entirely on device. Notable for being a one off licence rather than a subscription, which over a few years is a different kind of decision from everything else here.
- VoiceInk: Mac and iOS, open source, with the code on GitHub for anyone to read. Local transcription needs an Apple Silicon Mac. One off pricing from $29 for a single device.
The one you already own
Before paying for anything, run the numbers above using the dictation already built into your machine, because you have had it for years and it costs nothing. On a Mac or iPhone that is Apple Dictation. On Windows 11 it is voice typing, which you start with the Windows key and H. In Google Docs it is Voice typing under the Tools menu, and on an Android phone it is the microphone on the Gboard keyboard.
None of these do the language model cleanup pass, so expect the dictation cleanup figure to be higher than it would be with a paid tool. That is exactly the gap the paid tools are selling, and the calculator above will price it for you rather than making you take anyone's word for it. If a free tool already gets you most of the saving, that is worth knowing before a subscription, and nobody selling a subscription is going to tell you.
What happened to Dragon
Dragon is the name most people still reach for, and it is worth knowing what it is today, because a lot of the roundups you will find were written when the answer was different. Dragon Anywhere Mobile, the phone app, stopped taking subscriptions on 1 July 2026. Nuance says so plainly on its own Google Play listing: subscriptions to Dragon Anywhere Mobile are no longer available for purchase. If you have seen it recommended recently, that is why the link goes nowhere useful.
Dragon Professional v16, the Windows desktop product, is still sold and still very good at what it does, particularly custom vocabulary for specialist work. Two things to know before you compare it with the rest of this list. It is Windows only, so Mac users have no Dragon option at any price. And Nuance does not publish a price on the product page at all, it asks you to get in touch, while resellers list it around several hundred dollars as a one off licence. That is a different purchase from an eight dollar subscription, and the calculator above is the honest way to decide whether the gap is worth it to you.
Three things this calculator cannot know
Thinking is not typing. This is the big one. A typing test measures how fast you copy text that is already in front of you, and real writing is mostly deciding what to say. Dictation speeds up the words, not the deciding. If half of your writing hours go on composing rather than transcribing, halve every answer above. Some people find that talking helps them think, and some find it stops them thinking at all, and no calculator can tell you which one you are.
Recognition is not equally good for everybody. Speech recognition has repeatedly been found to be measurably less accurate for some accents and some speakers than for others. That means the editing overhead here is not one number for the whole population, and the published averages quietly hide the spread. If these tools work worse for you than the reviews promise, that is a documented pattern and not a failure on your part.
Dictation is loud. An open-plan office, a shared house, a coffee shop, a sleeping baby, a call you did not want overheard. Plenty of people for whom the arithmetic works beautifully still cannot use it for most of their day, and that has nothing to do with words per minute.