Whisper is the reason a whole generation of dictation apps exists. OpenAI released the speech model as open source in September 2022, under the MIT license, and developers have been building apps around it ever since. The model itself has no window and no hotkey, so the app you pick decides almost everything you will notice: whether text lands where your cursor is, whether it can chew through a two-hour interview, and whether your audio ever leaves your computer.
The short answer: the best Whisper apps in 2026 are MacWhisper for transcribing audio files on a Mac, Superwhisper for polished dictation on Mac and Windows, VoiceInk for open-source dictation on a Mac, Handy for free dictation on Windows, Mac and Linux, and Buzz for free file transcription on any computer. All of them can run Whisper locally.
We checked every price, platform and model detail below on each developer's own site or code repository on November 2, 2026. Prices change, so read them as "at the time of writing." We make Aqua Voice, which is not a Whisper app. It gets its own section near the end, after the picks.
What makes something a "Whisper app"
Whisper is a speech recognition model. OpenAI trained it on 680,000 hours of multilingual and multitask audio, according to its research paper, and it can transcribe speech in many languages, translate speech into English, and identify which language is being spoken. A Whisper app is any program that runs that model for you. Many of them use whisper.cpp, an open-source C/C++ port that runs on Mac, Windows, Linux, iOS and Android, with Apple Silicon support through Metal and Core ML.
Three questions sort the field quickly:
- Which job? Some apps transcribe files you already have: interviews, lectures, recorded meetings. Others dictate: you press a shortcut, talk, and the text appears in whatever app you are typing in. A few do both, but each is built around one.
- Local or cloud? Every app below can run Whisper on your own computer, so your audio never has to leave it. Several also let you connect a cloud provider when you want speed on a lighter machine.
- Which computer? Local Whisper is fastest with hardware acceleration: Apple Silicon on a Mac, a capable graphics card on a PC. Older machines still work with smaller models.
The OpenAI API change to know about
On August 26, 2026, OpenAI told developers that its hosted whisper-1 model will be removed from the API on February 26, 2027, and named gpt-transcribe and gpt-live-transcribe as the replacements. That matters for apps and scripts that send your audio to OpenAI's API under the Whisper name. It does not touch apps that run the open-source model on your own computer, which every pick below can do.
Which Whisper model should you choose?
Most Whisper apps ask you to download a model the first time you use them. OpenAI publishes six sizes, and the tradeoff is simple: bigger is more accurate and slower.
Model | Parameters | Memory needed | Speed vs large |
|---|---|---|---|
tiny | 39M | about 1 GB | about 10x |
base | 74M | about 1 GB | about 7x |
small | 244M | about 2 GB | about 4x |
medium | 769M | about 5 GB | about 2x |
large | 1,550M | about 10 GB | 1x |
turbo | 809M | about 6 GB | about 8x |
Three notes from OpenAI's own documentation. The English-only versions (tiny.en through medium.en) tend to do better on English, especially at the tiny and base sizes. Turbo is an optimized version of large-v3 that is much faster with a minimal drop in accuracy, but it is not trained for translation. And OpenAI measured those speeds on a data-center GPU, so your laptop will be slower; the ratios are what matter.
A reasonable starting point: turbo on a recent Apple Silicon Mac or a PC with a good graphics card, small or base on an older laptop, and large only when accuracy on a long recording matters more than waiting.
The best Whisper apps at a glance
App | Best for | Price at the time of writing | Platforms |
|---|---|---|---|
MacWhisper | Transcribing audio files | Free version; Pro about €65 one-time | Mac (separate iPhone and iPad app) |
Superwhisper | Polished dictation | Free tier; Pro $8.49/mo, $84.99/yr or $249.99 lifetime | Mac, Windows, iPhone, Android |
VoiceInk | Open-source dictation on a Mac | From $25 one-time; free to build from source | Mac (Apple Silicon), iPhone |
Handy | Free dictation, Windows included | Free | Windows, Mac, Linux |
Spokenly | Free local models on many platforms | Free; Local+ $29 one-time; Pro $12/mo billed annually | Mac, Windows, Linux, iPhone |
VoiceTypr | One-time price on Mac and Windows | $69 one-time for 2 devices | Mac, Windows |
OpenWhispr | Open source with an optional cloud | Free; Pro $80/yr | Mac, Windows, Linux |
Buzz | Free file transcription | Free | Mac (Apple Silicon), Windows, Linux |
Best Whisper apps for Mac
1. MacWhisper, best for transcribing audio files
MacWhisper, from independent developer Jordi Bruin, is built around one job: turning recordings into text. Drag in an audio or video file and you get a timestamped transcript you can search, edit and export as subtitles (SRT, VTT) or text. Transcription runs on your Mac unless you choose one of the optional cloud providers, which suits sensitive audio like interviews.
It supports the Whisper models from tiny to large-v3, plus NVIDIA Parakeet and Qwen models. The free version transcribes files and your microphone in 100 languages, exports subtitles, and includes basic system-wide dictation. Pro adds speaker recognition, batch and YouTube transcription, meeting recording in Zoom, Teams and other call apps, more export formats, higher-quality dictation and AI cleanup.
- Price at the time of writing: free, or Pro as a one-time license (€64 on macwhisper.com and €65 at the developer's Gumroad checkout when we checked).
- Platforms: Mac on macOS 14 or later (best on Apple Silicon, also runs on Intel). An iPhone and iPad version is sold separately in the App Store as Whisper Transcription.
- Worth knowing: the developer warns about copycat sites, so download it from macwhisper.com or the developer's Gumroad page.
- The honest tradeoff: dictation is a feature here, not the core. If you mainly want to talk into other apps, the next two picks are built for it.
2. Superwhisper, best polished Whisper dictation
Superwhisper is, in our view, the most polished of the Whisper dictation apps. Press a shortcut, speak, and the text is pasted into the app you are using, cleaned up by "modes" that set tone and format for messages, emails or code. Its free tier includes unlimited use of Whisper models (which run offline), voice-to-text in any app and support for more than 100 languages, which is generous.
Pro adds cloud speech and language models, your own API keys, translation into English and file transcription.
- Price at the time of writing: free tier; Pro at $8.49/mo, $84.99 a year, or $249.99 for a lifetime license, with a 3,000-word trial of Pro features and a 30-day refund.
- Platforms: Mac, Windows, iPhone and Android.
- The honest tradeoff: Superwhisper's own FAQ says offline models only run really well on Apple Silicon Macs and recommends cloud models on Intel Macs. We compare it with Aqua in Aqua Voice vs Superwhisper, and our Superwhisper alternatives roundup covers the field around it.
3. VoiceInk, best open-source dictation on a Mac
VoiceInk does Superwhisper's job with the source code in the open. It runs Whisper through whisper.cpp, and NVIDIA's Parakeet, entirely on your Mac; the developer says your voice never leaves the machine, and an optional cloud text-cleanup step sends only the transcribed text. The code is public under the GPL v3 license, so you can read it, audit it, or build it yourself for free.
- Price at the time of writing: $25 one-time for one Mac, $39 for two, $49 for three, with a free trial and a 14-day refund. Building from source is free.
- Platforms: Apple Silicon Macs on macOS 14.4 or later, plus an iPhone app.
- The honest tradeoff: it is Mac and iPhone only, and open source means the community is most of your support desk. See Aqua Voice vs VoiceInk.
Best Whisper apps for Windows
4. Handy, best free Whisper dictation on Windows
Handy is the answer if you want Whisper dictation on Windows and do not want to pay anything. It is free, open source under the MIT license, and runs completely offline on Windows, Mac and Linux. Hold a shortcut to record and release to stop (or tap to toggle), and Handy pastes the text into whatever you are typing in.
It offers Whisper small, medium, turbo and large with GPU acceleration when your computer has it, plus NVIDIA's Parakeet V3, which is tuned for ordinary processors and detects the language automatically.
- Price at the time of writing: free (the project is funded by sponsors and donations).
- Platforms: Windows, Mac and Linux. On Windows you can also install it with
winget install cjpais.Handy. - The honest tradeoff: it does one job, transcribe and paste, and it is a small open-source project with no phone app. Our Aqua Voice vs Handy page compares the two approaches.
5. Spokenly, best free local models on the most platforms
Spokenly runs local Whisper and Parakeet models for free with no word caps and no account, and it does so on Mac, Windows, Linux and iPhone. It also lets you connect your own API keys from cloud providers at no extra charge, so you can switch between offline privacy and cloud speed.
- Price at the time of writing: free for local models; Local+ is $29 one-time for extras like settings sync; Pro is $12/mo billed annually ($144 a year) and adds Spokenly's own cloud models and AI cleanup.
- Platforms: Mac, Windows, Linux and iPhone.
- The honest tradeoff: the free tier is the local engine; the cloud models that do the heavy lifting on slower machines are paid or bring-your-own-key. See Aqua Voice vs Spokenly.
6. VoiceTypr, best one-time price on Mac and Windows
VoiceTypr is for people who want one payment and both platforms. It uses on-device Whisper and Parakeet by default, uses your PC's graphics chip on Windows when it can, and lets you switch to cloud engines from Soniox, OpenAI, Groq or Deepgram on lighter machines.
- Price at the time of writing: $69 one-time for two devices (one- and four-device licenses are also offered), after a 3-day free trial with no card, and a 7-day money-back guarantee.
- Platforms: Mac on macOS 14 or later and Windows 10 or later.
- The honest tradeoff: the developer says the Windows installer is not code-signed yet, so Windows SmartScreen warns about an unknown publisher on install. It is a solo-founder product, so weigh that against the low price.
7. OpenWhispr, best open source with an optional cloud
OpenWhispr is open source under the MIT license and runs on Mac, Windows and Linux. Local dictation with Whisper (through whisper.cpp) or NVIDIA Parakeet is unlimited and free, with GPU acceleration on Metal, CUDA and Vulkan, and you can plug in your own cloud API keys or use OpenWhispr's own cloud.
- Price at the time of writing: free, including 2,000 words a week of OpenWhispr's cloud transcription; Pro is $80 per user per year and adds unlimited cloud transcription and an iPhone and iPad app.
- Platforms: Mac, Windows and Linux (iPhone and iPad on paid plans).
- The honest tradeoff: it also bundles meeting notes and an AI agent, so it is a bigger app than a plain dictation tool. See Aqua Voice vs OpenWhispr.
Best free Whisper app for transcribing files on any computer
8. Buzz
Buzz is the free, open-source counterpart to MacWhisper, and it runs on Windows and Linux too. It transcribes and translates audio and video files (or YouTube links) offline, can transcribe live from your microphone, identifies speakers, and exports TXT, SRT and VTT. It supports several Whisper engines, including whisper.cpp and Faster Whisper, with acceleration on NVIDIA graphics cards.
- Price at the time of writing: free, under the MIT license. The developer also lists a separate Mac App Store version.
- Platforms: Windows, Linux and Apple Silicon Macs (the last version for Intel Macs is 1.4.5).
- The honest tradeoff: Buzz is for files and live captions, not for typing into other apps. Its Windows installer is unsigned, so expect a warning on install.
For developers: run Whisper yourself
If you are comfortable in a terminal, you can skip the apps. OpenAI's own package installs with pip install -U openai-whisper (it needs ffmpeg), and one command transcribes a file:
whisper interview.mp3 --model turbo
For a faster build with no dependencies, whisper.cpp is the engine several of the apps above build on. And if you are moving a product off OpenAI's hosted whisper-1 before February 2027, our speech to text API roundup compares the hosted options, including our own Avalon API at $0.39 per audio hour through OpenAI-compatible routes.
Is Wispr Flow a Whisper app?
If you searched for "whisper flow," you probably meant Wispr Flow, a different product with a sound-alike name. It is not on this list because Wispr does not present it as a Whisper app. In 2026 the company introduced Canto, which it describes as its own speech model for real-time dictation. We compare it in Wispr Flow alternatives and Aqua Voice vs Wispr Flow.
Where Whisper falls short, and where Aqua fits
Whisper is a remarkable general-purpose model, and the apps above make it free or cheap to run privately. Its limits show up in three places:
- Hard words. Names, product terms and technical jargon are where general models slip. On AISpeak-10, a challenging subset of AISpeak, our own benchmark of real-world clips in which people say AI and coding terms, Whisper Large v3 got the key term right in 65.1% of clips (self-reported).
- Your hardware sets the ceiling. The best Whisper models want a fast GPU and several gigabytes of memory. On a lighter laptop you trade accuracy for speed.
- You manage the model. Picking sizes, downloading updates and tuning settings is part of the deal.
Aqua Voice takes the other path. It is not a Whisper app: it runs on Avalon, our own proprietary speech model, in the cloud. On the same AISpeak-10 test, Avalon got the key term right in 97.4% of clips. That figure is self-reported and dates from Avalon's launch in August 2025, so weigh it as one data point. For checkable context, Avalon debuted as the top proprietary model on the public OpenASR leaderboard in October 2025 (sixth overall at the time), and in a 9to5Mac test the writer read the same passage into built-in Mac dictation and Aqua and counted 17 errors versus 1 (one passage and one speaker, so an anecdote rather than a benchmark).
What that looks like day to day: hold Fn on a Mac and talk, and by our own figures text starts appearing in under 50ms, with final text in about 450ms. You can dictate at up to 230 words per minute into any app, teach it names with a personal dictionary, and fix text by voice with Edit Mode. It runs on Mac, Windows and iPhone in 49 languages, and on Android, where it is a button above the keyboard you already use, in dozens. One subscription covers all four platforms. Aqua has a SOC 2 Type II audit report and is ISO 27001 certified.
- Price: free for 1,000 words, then Pro at $10/mo billed annually or $12/mo month to month. See pricing.
- The honest tradeoffs: Aqua needs a connection and processes your audio in the cloud, so if audio must never leave your computer, pick one of the local apps above. It is built for live dictation rather than batch-transcribing a folder of recordings (MacWhisper or Buzz do that well). And the free allowance is 1,000 words, not unlimited like Handy.
New to dictation? Our setup guide covers the first week, and Best Dictation Software and Best AI Dictation Apps for Windows cover the wider field.
How to pick in one minute
- Transcribing recordings on a Mac: MacWhisper.
- Transcribing recordings on Windows or Linux, free: Buzz.
- Dictating on a Mac: Superwhisper for polish, VoiceInk for open source and a one-time price.
- Dictating on Windows, free: Handy, or Spokenly if you also want an iPhone app.
- Dictating on Mac and Windows with one payment: VoiceTypr.
- Accuracy on jargon, on your phone too, and no models to manage: try Aqua Voice.
FAQ
Is Whisper free?
Yes. OpenAI released Whisper as open source under the MIT license, so the model costs nothing to run on your own computer, and several apps that wrap it (Handy, Buzz, the free tiers of MacWhisper, Superwhisper and Spokenly) are free too. OpenAI's hosted whisper-1 API is a separate, paid service that OpenAI will remove on February 26, 2027.
Does Whisper work offline?
Yes, when the app runs the model on your computer, which every app in this roundup can do. You download a model once, and after that transcription needs no internet connection. Cloud options in those apps are optional.
What is the best Whisper app for Windows?
For free dictation into any app, Handy. Spokenly is also free for local models and adds an iPhone app, VoiceTypr suits people who want one payment for Mac and Windows, and Superwhisper is the most polished paid option. For transcribing files on Windows, Buzz is free.
Is Wispr Flow the same as Whisper?
No. Whisper is OpenAI's open-source speech model. Wispr Flow is a separate dictation app from a company called Wispr, and despite the similar name it is not a Whisper app as Wispr presents it. Wispr describes its own speech model, Canto, as built for real-time dictation.
