Verity listens to any debate, call, stream or video and verifies claims the instant they're spoken — verdict, confidence and sources in an always-on-top overlay. Nothing leaves your machine.
A live truth layer for the conversations and media you consume. Verity verifies spoken claims in seconds — verdict, confidence and sources in a calm always-on-top overlay. Audio never leaves your machine.
Not a search box you query after the fact — a truth layer that runs alongside whatever you're already listening to.
Claims are checked seconds after they're spoken. You see the verdict while the conversation is still happening — not in a write-up the next day.
Transcription runs entirely on your device with on-device Whisper. Your raw audio never touches the cloud — only the short extracted claim text is checked.
Your microphone and system audio — what you're watching or listening to. Every voice in the room gets checked.
Confidence score, a plain-English explanation and linked sources — never a bare true/false you have to trust blindly.
A glanceable glass panel that floats over anything — video, docs, your IDE — and stays out of the way until you need it.
Capture your mic, system audio, or both at once with a single click.
On-device Whisper turns speech into a live transcript. The audio stays on your machine.
Factual claims are pulled out and checked against trusted sources in seconds.
A verdict card appears in the overlay — verdict, confidence and sources, ready to glance at.
Catch spin in political debates and press conferences as it leaves the podium.
Verify numbers and claims live on sales, research and investor calls.
Keep creators and pundits honest while you watch or listen live.
Triage breaking-news claims with sources before you repeat or share them.
Transcription happens locally with on-device Whisper. The only thing that ever goes online is the short, extracted claim text — never a recording, never the raw stream. Nothing is stored or uploaded unless you choose to save it.
Verity never asks you to just trust it. Each card is built from four parts you can scrutinise yourself.
One of five — True, False, Misleading, Needs context, Unverifiable. Color and icon, never hue alone.
How sure the model is, shown as a meter — so a 71% reads differently from a 96%.
Why the claim got that verdict — the nuance a true/false label always misses.
The receipts — primary sources you can open and judge for yourself.
Free to download and free to start. macOS and Windows today; mobile is on the way.
For trying it out and the occasional check.
For anyone who checks claims every day.
For power users who want the sharpest accuracy.
For newsrooms, research and comms teams.