Brandon — Offline Speech-to-Text for Windows

Talk. Beep. Text.
Press CTRL + SHIFT + SPACE or say “Hey Brandon” to dictate anywhere on Windows. Everything runs fully offline on your PC—nothing sent to the cloud.

Why Brandon?

• Super simple. Brandon does just what it’s supposed to. Minimal. No Bloat. The App itself is 5 MB the rest is the model you choose.
• Always offline, privacy-first. Your audio never leaves your machine. Powered locally by whisper.cpp (a C/C++ port of OpenAI’s Whisper). (GitHub)
• Wake word or hotkey. Say “Hey Brandon” (starts listening ~1s later) or press Ctrl + Shift + Space. A quick beep confirms start/stop.
• Works anywhere in any app on your machine.
• Copies to your clipboard. Finish dictating, press the hotkey again, and your text is copied to the text field automatically, and saved to the clipboard if you need it in more than one place. (Tip: turn on Windows Clipboard history with Win + V if you want to keep past transcriptions.)
• No fluff. Minimal tray app, no accounts, no telemetry.

How it works

1. Start dictation — say “Hey Brandon” or press Ctrl + Shift + Space → you’ll hear a beep.

1. Speak normally. Use it in any app: notes, email, IDEs, chats.
2. Stop & copy — press Ctrl + Shift + Space again. Brandon transcribes and copy pastes the results to the text field and saves to your clipboard for pasting again later if needed.

Fast installation, simple choice of models (which you can change later, just re-run the installer it will download the model from the official repository, be sure to delete models you don't use anymore) you can switch between models by changing the path in the settings.

Models & languages

Choose English-only models for speed, or multilingual Whisper models for many languages. Smaller models = faster; larger models = higher accuracy, I use the small model for english personally. To write this in fact and it works just fine.

I almost only use the base model, it's super fast and accurate for english, and as you can tell I have an accent but that isn't an issue. I run it on the CPU and it clocks in at about 300 MB Memory during usage.
The App itself takes up 5.1 MB.
The Multilingual Models vary in size and Memory but can need up to 3 GB. See the official Repo here, where you can also see if other models are available.
You can always also just use your own models. Simply set the path in the settings.

What you get

• Windows installer (Minimal 2.5 MB, requires Windows VC++ which almost everyone should have, otherwise use the static installer that bundles it)
• Brandon app with wake word + hotkey dictation
• Clipboard-first workflow (Brandon itself doesn’t keep a history; your system clipboard can, if enabled)

Requirements:

The Base Model needs about 300 MB to run, all models run perfectly fine on my CPU (12 Core i5 10th generation).
Recommend is at least a CPU with 4-8 Cores, 8 GB of memory. And for the models you'll need between 150 to 3 GB of disk space, which you can see during the installation.

Requirements

• A Windows PC and a microphone.
• Runs completely offline; no internet required.

Price & source

$5 — a modest way to support this build. I’ll open-source Brandon later so you can extend or roll your own if you prefer free.

FAQ

Does Brandon paste automatically?
Yes. When you stop dictation, Brandon auto-pastes into the active text field and the transcription remains on your clipboard. If you use Windows’ clipboard history, you can still pull it up with Win + V.

Is anything uploaded?
No. All processing happens locally on your computer through whisper.cpp—no audio or text is sent anywhere. (GitHub)

Which model should I pick?
If you want speed on modest hardware, try an English-only small model; for more languages or accuracy, pick a larger multilingual Whisper model.

Does it work in every app?
If the app accepts typed text or paste, Brandon will work there.

Support?
Message me directly on Gumroad (email there is fine).

Get Brandon for $5 → Install, say “Hey Brandon,” and start dictating offline in seconds.

Notes & credits

• Built on whisper.cpp and OpenAI’s Whisper research (multilingual ASR trained on a large, diverse dataset). (GitHub)
• Installer packaged with Inno Setup, a long-trusted Windows installer. (JR Software)
•