The best dictation software for most people in 2026 is superwhisper - it runs entirely on your machine, costs $84.99 a year or $249.99 once, and never sends your audio anywhere. If you want the polish of a cloud service and don't mind the round trip, Wispr Flow is the better-finished product at $12 a user per month.
Those are the two-line answers. The longer answer is that "dictation software" covers at least five different jobs, and the tool that wins depends entirely on which one you're doing. Typing at the cursor in any app is not the same product as transcribing a recorded interview, which is not the same product as writing code by voice.
This guide covers all of them. Every price below was checked against the vendor's own pricing page, and every tool has a full profile in the directory.
| Tool | Best for | Platforms | Price | Offline |
|---|---|---|---|---|
| superwhisper | Most people | macOS, Windows, iOS | Free; $84.99/yr or $249.99 lifetime | Yes |
| Wispr Flow | Polish, mobile included | macOS, Windows, iOS, Android | Free tier; $12/user/mo annual | No |
| MacWhisper | Transcribing files on a Mac | macOS | Free tier; €59 one-time | Yes |
| Handy | Free and offline, any OS | macOS, Windows, Linux | Free (MIT) | Yes |
| VoiceInk | Cheapest good Mac app | macOS (Apple Silicon) | $29 one-time | Yes |
| Apple Dictation | Spending nothing at all | macOS, iOS | Free, built in | Yes |
| Aqua Voice | Lowest latency | macOS, Windows, iOS | Free; $8/mo annual | No |
| Voice In | Dictating into web apps | Chrome, Edge | Free tier; $60/yr | In-browser |
| Otter.ai | Meeting notes | Web, iOS, Android | Free; $8.33/mo annual | No |
| Fathom | Free meeting recording | Web | Free; $16/user/mo annual | No |
| Freed | Solo clinicians | Web, iOS, Android | $39–119/mo | No |
| Talon Voice | Coding by voice | macOS, Windows, Linux | Free | Yes |
Most bad dictation purchases come from buying the wrong category, not the wrong brand. There are five, and they barely overlap.
Dictation at the cursor. You press a key, talk, and text appears wherever you were typing - an email, a Slack message, a code comment. This is what most people mean by dictation software. Browse General Dictation.
File transcription. You have a recording and you want a document. Different constraints entirely: batch processing, speaker labels, subtitle export, and no need for low latency. MacWhisper, Buzz, and Vibe do this for free, offline.
Meeting assistants. A bot joins your call, or the app listens to your machine's audio, and you get a summary and action items afterwards. See Meeting Transcription.
Clinical and legal dictation. Specialist vocabulary, EHR integration, and a signed Business Associate Agreement. Nothing in the general category is a substitute. See Medical & Legal.
Voice coding. Mapping speech to editor commands rather than prose. See Voice Coding.

superwhisper runs Whisper models locally on macOS, Windows, and iOS. The free tier gives you unlimited dictation on the small models, which is genuinely usable. Pro is $8.49 a month, $84.99 a year, or $249.99 once - and the lifetime option is the reason a lot of people land here, because at roughly two years of a competitor's subscription it stops being a recurring cost.
What you get for that is speed that doesn't depend on your connection, custom "modes" that change the output format depending on what you're doing, and the option to route to cloud models with your own API keys if you want a stronger model for a specific job.
The trade-off is configuration. superwhisper expects you to pick models, set up modes, and tune things. If you want an app that works well the moment you install it and never asks you a question, that's Wispr Flow or VoiceInk.

Wispr Flow is the most finished product in the category. It strips filler words, adjusts register between a formal email and a Slack reply, learns your vocabulary, and ships on macOS, Windows, iOS, and Android - the only mainstream option that covers phones properly.
Free gives you 2,000 words a week on desktop. Pro is $12 per user per month billed annually, $15 monthly.
The catch is structural: every dictation is a round trip to Wispr's servers. There is no on-device mode. It won't work on a plane, and it's a non-starter under a policy that keeps audio on the machine. Privacy Mode limits retention, not transmission. Users also report a one-to-two second wait in practice against a published sub-700ms p99, and the Windows build attracts most of the reliability complaints.

Three answers, depending on what "free" needs to cover.
Apple Dictation is already installed on your Mac and iPhone, runs fully on-device on Apple Silicon, and in 2026 added automatic capitalisation and punctuation. Try it before you pay for anything - a meaningful share of people who buy a dictation app never tested the one they already had.
Handy is MIT-licensed, free, offline, and treats Linux as a first-class target rather than an afterthought. It runs Whisper Small through Large with GPU acceleration, plus Parakeet V3.
Whispering is free in the browser or as a desktop app, with local GGUF models on desktop and bring-your-own-key cloud models if you want them.
We go deeper in the best free dictation software.
If audio must not leave the machine - a compliance rule, a client contract, or a personal preference - the field narrows fast. superwhisper, VoiceInk ($29 one-time, Apple Silicon), Handy, Parrot (free, Mac, no account required), and Dragon Professional all process locally.
Watch for the two words vendors use to blur this: "on-device" sometimes means only the wake-word detection is local, and "private" often describes a retention policy rather than where the processing happens. The works offline collection filters to tools that actually run without a connection.

Meeting tools price on three different axes, and the free tiers are where the differences show.
| Tool | Free tier | Paid from |
|---|---|---|
| Fathom | Unlimited recording, 5 AI summaries/mo | $16/user/mo annual |
| Fireflies.ai | Unlimited transcription and summaries | $10/seat/mo annual |
| Otter.ai | 300 min/mo, 30 min per conversation | $8.33/mo annual |
| tl;dv | Unlimited recording, 10 AI uses/mo | $18/user/mo annual |
| Granola | Unlimited meetings, limited history | $14/user/mo |
Fathom has the most generous free recording. Fireflies.ai is the only one that gives you unlimited transcription and unlimited summaries at $0, metering a separate pool of AI credits instead. Granola is the pick if you don't want a bot visibly joining the call - it captures device audio instead.
One thing to know before you deploy any of them: Otter.ai is facing a class action over recording consent. Whatever tool you choose, check your jurisdiction's rules on recording other people.
Full breakdown in the best AI meeting notetakers.

Nothing in the general category is appropriate for patient encounters, regardless of what its marketing says about HIPAA. You need a signed BAA, and you need it to cover your specific plan.
Freed is the one an individual clinician can actually buy: $39 a month for 40 notes, $79 for unlimited, no procurement process. Heidi Health has a free tier with unlimited basic consults. Dragon Medical One remains the incumbent at roughly $99 per user per month plus about $525 in one-time implementation. Abridge has been Best in KLAS two years running but sells only to health systems.
A warning on free tiers here: Nabla's free plan reportedly comes without a BAA. Free and HIPAA-compliant are not the same thing. See the HIPAA-compliant collection and our guide to HIPAA-compliant transcription.

Talon Voice is the foundation almost everything else is built on: free, local, scriptable in Python, with voice commands, noise recognition, and eye tracking. Add Cursorless on top if you use VS Code and you get whole edits in a single command, at the cost of weeks of learning on top of Talon's own curve.
Serenade is friendlier to start with, but the original project is unmaintained and development has moved to a community fork. More in the best voice coding tools.
Dragon Professional for new buyers. $699.99, Windows only, on a product line being wound down - Dragon for Mac died in 2018, Dragon Home in 2023, Dragon Anywhere mobile in July 2026. It's still excellent at what it does, and it's still the accuracy benchmark with a trained profile, but you're buying into a shrinking platform. See Dragon alternatives.
Speechify if you want dictation. It's a text-to-speech product that added voice typing later, and voice typing is Premium-only at $29 a month.
Windows Speech Recognition. Deprecated in December 2023 and removed from current Windows 11 builds. Use Win+H voice typing or Voice Access instead.
Lifetime deals from very new vendors. VoiceDash sells through AppSumo; the 60-day guarantee is real, but a lifetime licence is only worth what the company's remaining lifespan makes it.
Every tool in the directory is listed with the platforms it runs on, how it's priced, and whether it works offline - the three filters that eliminate most options before subjective quality matters. We don't take payment for placement, and pricing is checked against vendor pages rather than review-site summaries.
Where a vendor's accuracy claim appears in this article, it's labelled as a vendor figure. Nobody in this industry publishes independently audited word error rates, and the numbers in marketing copy are measured on clean audio in quiet rooms. See speech-to-text accuracy explained for what those percentages actually mean.
On clean audio in a quiet room, the large Whisper-family models used by superwhisper, MacWhisper, and Handy are at or near the top, and Dragon Professional with a trained profile and specialist vocabulary is still hard to beat on domain terms. Accuracy differences between good tools are usually smaller than the difference a decent microphone makes.
Often, yes. Apple Dictation is free, offline, and fine for most short-form writing. Handy and Whispering run the same Whisper models the paid apps use. What you pay for is polish: filler-word removal, custom vocabularies, per-app formatting, and support.
Some does. Anything running Whisper or a similar model locally - superwhisper, VoiceInk, Handy, Buzz, Parrot - works with no connection at all. Cloud tools like Wispr Flow, Aqua Voice, and every meeting assistant do not.
The market has settled around $8–15 a month for cloud dictation, $29–250 one-time for local Mac apps, and $0 for open-source. Meeting assistants run $10–30 per seat per month, and clinical scribes $39–500 per clinician per month. See our pricing guide.
Only with a tool that will sign a Business Associate Agreement covering your plan. That means Freed, Dragon Medical One, Suki, Abridge, or similar - not a general dictation app with "HIPAA" in its marketing copy.
Last verified: 10 August 2026.