Setup & help

1 · Install

1. Download Pipevoice-Setup.exe and run it. No admin rights needed — it installs just for you.
2. Windows will show a blue "Windows protected your PC" screen. Click More info → Run anyway. This appears for every app without a paid code-signing certificate; PipeVoice is open source, so you can read the whole thing first.
3. It starts quietly in your system tray. Look for the mic icon near the clock — that is the whole interface. Right-click it for everything.

Updates install themselves in the background. You can switch to early builds under Settings → Automatic updates → Get updates early (beta).

2 · Choose an engine

On first run PipeVoice asks how you want your speech turned into text. Every option is either free or costs a few pence a day, and you can change it any time from the tray menu.

Getting a key: Gemini → aistudio.google.com/apikey · Groq → console.groq.com/keys · Deepgram → console.deepgram.com. Paste it into the first-run box, or add it later under Settings → API keys.

3 · Start talking

Hold Ctrl + \, speak, let go. Your words type themselves in wherever the cursor is — any app, no integration needed. A small pill in the corner shows a live mic meter while it listens.

The second hotkey

Hold Right Ctrl + Right Shift instead and the text goes to your clipboard rather than being typed. Useful when the app you are in does odd things with keystrokes.

Hold, or tap?

• Push-to-talk (default) — hold while speaking, release to type. Best for short bursts.
• Toggle — tap to start, tap to stop. Best for long dictation. Switch in Settings.

Make it yours

Right-click the tray icon → Settings. Rebind either hotkey, choose your microphone, switch typing to pasting, set a language, and add Vocabulary — names and jargon it keeps getting wrong. Everything applies immediately; there is no restart.

4 · Record a meeting

PipeVoice can capture both sides of a call — your mic and whatever is coming out of your speakers — then transcribe it with who-said-what labels. It is off until you give it a hotkey.

1. Settings → Meetings, set a Meeting hotkey.
2. Press it to start. A pill appears in the corner; drag it anywhere you like.
3. Press again to stop. The recording lands in the Meetings tab, ready to transcribe.

PipeFocus (Deepgram only) reads the meeting live and nudges you with the occasional short tip. Off by default, in Settings.

Fix a problem

It says my API key is missing, but I definitely added one

This was a real bug, fixed in 2.35.3. A blank line in a leftover settings file could hide your actual key. Open the tray menu → About; if it says anything below 2.35.3, let the update install and it resolves itself — your existing key just starts working again. Nothing to re-enter.

My words aren't appearing in one particular app

If that app runs as administrator — plenty of terminals do — Windows refuses keystrokes from ordinary apps. Run PipeVoice as administrator too so the two match. Or use the clipboard hotkey and paste.

Nothing happens at all when I hold the key

• Check the hotkey has not been claimed by another app — rebind it in Settings and try something unusual like F9.
• Check an engine is selected and, for the cloud ones, that a key is saved (Settings → API keys shows Saved or Not set).
• Gaming and streaming overlays sometimes swallow global hotkeys. Close them and retest.

It's picking up the wrong microphone

Settings → Microphone. Blank means the Windows default. Pick one by name if you have several — headset, webcam and laptop mics are easy to mix up.

The first offline transcription takes forever

Local Whisper downloads its model the first time only (~150 MB). After that it is cached. Larger models are more accurate and slower; a GPU makes a large difference.

My claps or snaps aren't marking the recording

If you are on a Copilot+ PC, Windows Studio Effects is on by default and filters out exactly the sharp transient sounds the snap detector listens for. Turn it off in Windows Settings → System → Sound → your microphone.

A long recording stopped by itself

There is a safety cap so an unattended capture cannot fill your disk — 4 hours by default, adjustable in Settings → Meetings. PipeVoice also keeps only your newest 20 sessions and deletes older ones; raise that number if you need a longer archive.

Live transcription dropped mid-meeting

Deepgram closes an idle connection after a while. PipeVoice reconnects on its own and keeps recording throughout — the audio file is never at risk, only the live text. If it keeps happening on a stable connection, send us the log.

Where your files live

Everything is under one folder. Paste this into the File Explorer address bar:

Meeting recordings are kept apart from all of that, because they get large. By default:

Point Settings → Save meetings to at another drive if you record a lot. Existing recordings stay where they are — only new ones follow the new setting.

Report a bug

Open an issue on GitHub. Two things make it fixable in one round instead of five:

1. The version — tray menu → About.
2. The log — %APPDATA%\Pipevoice\pipevoice.log. Attach it, or paste the last 50 lines.