WindyWord
You were never meant to type your vision.
Speak it into existence.
Emails, documents, code, notes, text messages — on your Mac, Windows or Linux computer. On a Chromebook you can install it through the Linux section; we are still testing dictation there. No internet required in local mode. Never touch a keyboard to type again.
Life will never be the same.
Download Windy Word
Free. Private. Yours. Ten voice and translation models that run on your own machine — three in the installer, seven whenever you want them. A free Windy account is all you need to download.
v1.9.64.251 · macOS · Windows · Linux
Updating from an older version?
On a Mac, if Restart & update does nothing, quit Windy Word and open it again, or download it from windyword.ai.
First-time setup prepares the speech engine on your computer: a few minutes on most PCs, and up to an hour on a very slow one (2 GB of memory or a single slow core). Leave it running.
Speak it into existence
Breathe out.
Words appear.
Hold a key, say what you mean, let go. Windy Word writes it wherever your cursor is.
Let There Be Text
One Voice. Many Languages.
An illustration: dictate once, then translate the text into the languages you choose.
Waiting...
Waiting...
Waiting...
Waiting...
Waiting...
Your voice becomes text on your device. No internet. No upload. No cloud. Just instant voice-to-text.
The Six Powers
Your voice isn't just input. It's a force of creation.
The Keyboard Was an Accident
QWERTY exists to stop typewriter arms from jamming — a machine almost nobody alive has used. It is the grandmother who cut the ends off every roast because her pan was too small. We kept the cut. We lost the pan. A hundred and fifty years later you are still typing around a jam that cannot happen.
Get Your Head Back
Typing is not thinking. It is motor planning, finger sequencing, spelling, and watching your own hands — a stack of work running underneath the actual idea. Speaking skips all of it. The capacity you were spending on transcription goes back into the thought itself.
Think On Your Feet
Humans have strategized while moving for as long as there have been humans. It is well established that people think better walking than sitting. Put on a headset and dictate while you pace the room — most people find their ideas arrive faster and in better order than they ever did at a desk.
Your Words Never Leave
Not encrypted in transit. Not deleted after processing. Not sent at all. With a local model selected, your voice is transcribed on your own machine and the words never touch a network. Cloud dictation tools cannot say this — not as a policy, but structurally. Their product is the round trip.
Never Lose a Prompt Again
Everyone who dictates has lost one — you build the thought perfectly, and it dies somewhere between your microphone and a server. A dropped connection, a timeout, a bad hotel wifi. Local transcription has no round trip to fail. There is no connection to lose because there is no connection.
A Record of How You Think
Every word you dictate builds a corpus of you — what you worked on, decided, and cared about. Text is tiny; years of it is still a few megabytes. When your agent needs to catch up, it reads all of it in seconds and arrives as though it had been there the whole time. That corpus only exists if you start it.
Or just press WindyTune
10 models is a lot of choosing, so you don't have to. WindyTune reads what language you're speaking and how hard your machine is working, and picks the right engine for this moment. Bigger models hear more but weigh more; lighter models fly. You never have to know that — but you can always override it.
For most of history, exactly one group of people never had to write anything down. Pharaohs and magistrates spoke, and scribes, builders, and generals turned it into the world. Everyone else pushed their vision through a straw.
Speak. Let the world get to work.
Nobody is more passionate about the great art of voice-to-text than Windy Word, Windy Translate and Windy Traveler.
Our lab forks from strong open foundations and refines them in-house, and keeps refining them. These are our models. We are on a permanent quest for the lightest, fastest, most accurate voice models anywhere in the world — and every one of them runs on your machine, not ours.
Open the Vault
10 models. 3 in the box, 7 a click away. All of them free, all of them yours, all of them running on your machine.
We fork them, fine-tune them, quantize them, and test and document every one in our own clinic. They are ours. Download them, delete them, download them again — as often as you like, forever, at no cost.
In the box
3 engines · 0.37 GB · ready the moment you installWe keep the installer light on purpose. These cover most people on day one — no downloads, no setup, no decisions.
Hears any of 99 languages and writes them down in the language you spoke.
Hears any of 99 languages and writes them down in the language you spoke.
Hears any of 99 languages and writes them down in the language you spoke.
Download when you want them
4 more voice enginesBigger models hear more and cost more machine. Take them when you want them, delete them when you don't.
Built for English and nothing else — which is exactly why it is fast. If you dictate in English, this is the one.
Hears any of 99 languages and writes them down in the language you spoke.
Hears any of 99 languages and writes them down in the language you spoke.
Hears any of 99 languages and writes them down in the language you spoke.
Do you even need a translator?
Probably not. Your voice engines already do two things on their own, with nothing downloaded:
- • Speak Spanish, get Spanish — they write down about 99 languages in their own language.
- • Speak Spanish, get English — they can turn any of those 99 into English.
You only need a translator for the third case: one non-English language into another — Spanish → French, German → Italian.
The translators
3 models · any language to any otherPick by what your machine can carry. Every one is permissively licensed, so anything you make with them is yours to sell.
Good on widely-spoken languages, weaker on rarer ones. The one to pick if your computer is older or has 8 GB of memory or less.
More accurate than Nano and holds up better outside the most common languages. Comfortable on a computer from the last five years or so.
The most accurate of the three and the only one covering 452 languages and dialects. It is also 6× the size of Nano and works the hardest — on an older or low-memory machine it will be slow.
Not just languages — dialects
Canadian French is not France French. Traditional Chinese is not Simplified. Windy Translate Max knows the difference on 33 of them.
Translation quality varies for low-resource languages. We publish the count honestly rather than promising every one of 452 is equally good.
The Ritual
Three steps. No complexity. No ceremony required.
Summon
Create your free account and install. 3 engines come in the box, so there is nothing to choose and nothing to download before you can start.
macOS · Windows · Linux
Speak
Just talk. Speak naturally and your words become text — in the language you spoke. WindyTune picks the right engine for your machine on its own. No configuration, no menus, no friction.
99 languages · auto-detect
Create
Paste your text wherever your cursor is blinking — emails, documents, code editors, chat apps. Your spoken words, converted to text, ready to use. No typing. Ever.
Paste anywhere · No typing · No internet needed
Ready for more? Storage and sync in WindyCloud are coming soon. Your voice can become a voice clone with WindyClone, or power real-time conversations in WindyChat.
Built for Real Work
Where never touching a keyboard actually changes the day.
Everyone speaks. Now everyone creates.
Dictate the visit while it is fresh, in the room, without a cloud service in the middle of protected health information. The words are transcribed on the machine in front of you and go nowhere else.
Reporters, field researchers, and anyone working past the edge of coverage. The engines run offline, so a dead connection costs you nothing — there was never a round trip to lose.
Chapters, briefs, design docs, and the sprawling prompts nobody wants to type twice. Pace the room with a headset on and let the draft arrive at the speed you actually think.
Speak any of about 99 languages and get it back in that language, or in English. Switch mid-sentence — the app works out what you are speaking without being told.
Join the Movement — Speak Your First Word
Claim Your VoiceThe Offering
Free is in the budget, baby.
Windy Word
Not a trial. Not a teaser. The whole thing.
- ✓All 10 models — 3 in the installer, 7 whenever you want them
- ✓Unlimited dictation — talk for two hours if you like
- ✓99 languages of speech, auto-detected — you never pick your input language (Windy Edge, our English specialist, hears English only)
- ✓Translation into as many as 452 languages and dialects
- ✓Runs entirely on your machine — works in airplane mode after your first sign-in
- ✓Delete any model and re-download it later, as often as you like
- ✓Commercial use allowed — every model is permissively licensed
- ✓WindyTune — the app picks the right engine for your machine
- ✓Cloud dictation up to 5 minutes, when your machine wants backup
Free account required. We ask for an email so you can recover your password.
Cloud plans, when your machine wants backup. Everything above stays free forever.
Pro
Cloud compute when your laptop is working too hard.
- ✓Everything in Free, forever
- ✓Cloud recordings up to 15 minutes
- ✓Model Lab (desktop app): choose which engine runs your cloud dictation
Windy Ultra
Longer cloud sessions.
- ✓Everything in Pro
- ✓Cloud recordings up to 30 minutes
Windy Max
The longest cloud sessions.
- ✓Everything in Ultra
- ✓Cloud recordings up to 60 minutes
Local voice-to-text stays free and unlimited on every plan, forever. Cancel anytime from your account.
Full refund within 14 days of your first charge · Refund policy
Your Voice Builds a Kingdom
WindyWord is the gate. Behind it: an empire where your voice becomes your most powerful asset.
"You just spoke 100 hours of meetings into existence. That voice data is GOLD. WindyClone turns it into your digital twin. WindyCloud is being built to keep it in one place for you. Coming to WindyChat: your twin speaking YOUR words in other languages. You're one spoken word away from owning your AI identity."
WindyCloud
The Gravity Well
You just transcribed 100 hours of audio. That data is gold. Windy Cloud is being built to keep everything you say, record and make in one place, for you and your agent, if you choose. Storage and sync for Windy Word are coming soon.
WindyClone
Your Digital Twin
Your voice data can create something extraordinary — a clone that speaks like you, looks like you, thinks like you. In any language. One click from the data you already have.
WindyChat
Chat, with translation coming
Messaging, voice notes and calls today. Coming: translated chat in your own voice — with your WindyClone, they hear YOUR voice speaking THEIR language.
WindyTraveler
Your AI Travel Companion
Regional translation packs for the road — download the languages of where you are going before you get on the plane. A separate app, sold separately, for travel.
Building on our models?
Windy Translate is the developer platform • windytranslate.com · soon
One voice. One kingdom. Speak and it shall be built.
Claim Your Voice
Retire your keyboard. Speak and your words become text — instantly, locally, no internet needed. Paste anywhere.
Free account · No credit card · No internet required after install · Just voice-to-text
macOS 11+ (signed + notarized by Apple) · Windows 10/11 · Linux x86_64 · 4GB RAM minimum
The Covenant
Not renting. Not licensing. Owning.
Your Words. Your Machine.
Your recordings are processed and stored on your own computer. Nothing is uploaded unless you choose it.
Browse Models →No Kill Switch. Ever.
If WindyWord disappeared tomorrow, your models still work. No kill switches. No phone-home requirements. No expiration.
Your Voice Is Your Temple
Local processing by default = your voice stays on your device. Cloud is opt-in, never automatic. We collect anonymous app and site events, never your audio or transcripts. We don't train on your audio.
Why We Give It Away
Because trust is earned, not demanded. You can inspect every model, verify every claim, and understand exactly what happens with your voice data.
Because freedom beats convenience. You're not renting access. You're not dependent on our uptime. You own your AI stack the way you own a book on your shelf.
Because the ecosystem is the business. WindyWord is free. WindyCloud, WindyClone, WindyChat — that's where we build a sustainable company. Your success with WindyWord is our best marketing.