Get Vlocalhost · on-device · latest
Install Vlocalhost
The offline voice engine that writes your meetings. Pick your platform, download one installer, and you're recording notes in a couple of minutes — nothing leaves your machine.
Detailed installation guide (PDF) — every platform, with troubleshooting ↓
When it doesn't go smoothly
Troubleshooting.
The five things people actually hit. The installation guide (PDF) covers every platform in detail, and the support page takes it from there.
“Windows protected your PC”
Expected — the installer isn't code-signed yet. Click More info → Run anyway. It's not a virus warning and says nothing about the file's contents; every line of what it installs is public.
“Apple could not verify”
Open System Settings → Privacy & Security, scroll to the line naming Vlocalhost, press Open Anyway. Once only. Right-click → Open no longer works on macOS 15+.
No microphone found
PortAudio is the one library the bundle can't carry, because it talks to your sound
server. sudo apt install libportaudio2, or dnf install portaudio.
Transcript saved, no summary
Ollama wasn't running. The transcript is written first on purpose, so it's never lost
to this. Open http://localhost:11434/api/tags — if that errors, start
Ollama, then Settings → Models → Re-check.
Nothing happens on click
The shortcut runs without a console, so the error had nowhere to go. Run the app from a terminal once and it will tell you what's wrong — the exact command per platform is in the PDF.
Where are my notes?
Outside the program folder, always — so updating or uninstalling can't touch them.
Run Vlocalhost --paths to print every location, and
--diagnose for a report to attach to a bug.
No model limitation
Bring your own voice model.
Vlocalhost doesn't lock you to one engine. Attach the speech-to-text model that fits your accent, language, and hardware — small and fast, or large and pin-sharp. Your model, your rules.
Any Whisper model
From tiny.en to large-v3, or any fine-tuned CTranslate2 model on Hugging Face. Pick it in Settings → Models.
A local model folder
Press Folder… in Settings and choose a converted model on disk — any size, any language, fully offline, no download.
Your own engine
Set CUSTOM_TRANSCRIBER to your own class with a .transcribe() method and bypass Whisper entirely.
