Dictate into any app
Email, WhatsApp, Word, a browser tab: hold two keys, or tap the bubble on Android with your own keyboard still there, and talk. Long messages are fine.
Berry is a voice assistant and dictation app for Windows and Android that runs entirely on your device. Ask it to open apps, set reminders and find files, or speak and watch your words appear in any app. No internet, no account, no subscription.
Free during early access · Windows 10/11 and Android 8+ · no account needed
What Berry does, how it works, and why your voice never has to leave your device.
Hold two keys and speak. Berry writes it out, with punctuation and without the “ums”, straight into the app you’re using. On your phone, its bubble floats above your own keyboard.
Three steps, all on your own processor. Nothing is uploaded, so it works the same in airplane mode.
On a PC, tap Alt + Win to ask Berry something, or hold Ctrl + Win and talk to type. On Android, say “Hey Berry” or tap the floating bubble above any keyboard.
A neural voice detector hears when you start and stop, then Whisper turns speech into text, right on your processor. Fillers like “um” vanish, punctuation appears, and your own names and words are spelled the way you like.
It opens apps, sets reminders and timers, finds files, calls and texts on your phone, answers questions with a local AI model, or types what you said into whatever app you’re in. Then it answers in a natural voice.
Everything Berry does runs locally. Online chat is there if you want a bigger model, and off until you turn it on.
Email, WhatsApp, Word, a browser tab: hold two keys, or tap the bubble on Android with your own keyboard still there, and talk. Long messages are fine.
Open apps, set reminders, timers and alarms, find files, call and text, control music and the torch.
Say the wake word and ask. Listening for it happens on your device and pauses when your screen is off.
Ask anything and a small AI model answers on your device. Want a bigger one? Switch on online chat.
Teach Berry names and terms once (“Ammi”, “Kubernetes”, a colleague’s name) and it gets them right.
Replies in a natural neural voice, with several to choose from, generated on your device too.
See what you dictated and the minutes you saved. After the days you choose, only word counts remain.
Berry sleeps until you need it and unloads its models when idle, so it stays cool and quiet.
Offline is the default. Add a free Groq key in Settings if you want a larger model for chat.
Wispr Flow is great at dictation, but it sends your voice to its servers and charges monthly. Siri and Google Assistant live on their own platforms. Berry does both jobs on the devices you already have, without uploading a word.
| Feature | Berry | Wispr Flow | Siri / Google Assistant |
|---|---|---|---|
| Works with no internet | Everything, always | Needs a connection; transcription runs in the cloud | Some requests, on recent devices |
| Your voice stays on your device | Audio is never uploaded | Sent to their servers to transcribe | Depends on the request |
| Dictate into any app | Windows and Android | Mac, Windows, iPhone, Android | Through the system keyboard |
| Opens apps, reminders, files | Built in | Dictation and text editing | On their own platform |
| Runs on a Windows PC | Windows 10 and 11 | Yes | Siri: Apple devices only |
| Price | Free during early access, no subscription | Free tier with a weekly word limit; Pro $12–15 a month | Included with the device |
Based on each product’s public pricing and help pages, November 2026. Features change; check their sites for the latest.
Straight from here: no store, no sign-up. Installing takes about two minutes.
Version 1.1.0 · 89 MB · Windows 10 or 11, 64-bit, 8 GB RAM recommended
Download Berry-Setup.exeVersion 1.1.1 · 94 MB · Android 8.0 or newer, 64-bit phone
Download Berry.apkFree while Berry is in early access. Later, new downloads will cost a small one-time price, never a subscription. Updates appear inside the app; download the new version from this page.
Yes. After the one-time model download, speech recognition, commands, chat and Berry’s voice all run on your device. Turn on airplane mode and try it. The only optional online parts are a once-a-day update check (which you can switch off) and online chat, if you add your own Groq key.
Windows SmartScreen warns about any new app that hasn’t been downloaded many times yet. Click “More info”, then “Run anyway”. Berry installs just for your user account and doesn’t need administrator rights.
Open the downloaded Berry.apk. Android asks once whether your browser may install apps; allow it, and the install continues. Berry asks for the microphone, and for accessibility only if you want the floating dictation bubble.
Most PCs from the last five years are fine; 8 GB of RAM is comfortable. If chat feels slow, pick the lighter chat model in Settings, or use online chat. On phones Berry needs a 64-bit Android 8 device; it unloads models when idle so it doesn’t drain the battery.
English works best, including Indian and Pakistani accents. Whisper also understands Urdu, Hindi and many more languages; choose yours in Settings.
Berry checks this site once a day for a newer version and shows “Update available”. Download it from this page and install over the old one; your settings and history stay.
It’s free during early access. Later, new downloads will cost a small one-time price. There will never be a subscription, and the version you have keeps working.