Local isn't a marketing word here
Debrief doesn't upload your calls anywhere to transcribe them. Recording, transcription, and filing all happen on your own machine, using an open source model, not a cloud API. Turn off your wifi mid-call and it keeps working, because it was never talking to anything but your own computer. The full source is on GitHub if you want to verify that yourself instead of taking our word for it.
Most "privacy-focused" call tools mean their servers follow a privacy policy. Debrief means there's no server in the loop at all. The distinction matters: a policy is a promise someone could break or quietly change, a structural fact isn't. Your audio goes from your microphone, through a local, open source Whisper model, into a plain text file on your own disk. Nothing in that chain is capable of leaving your machine, not because it's configured not to, but because it's never sent anywhere in the first place.
Audio moves through your machine only, mic to Whisper to a plain markdown file. It never crosses to a server, and cutting your wifi mid-call proves it: transcription keeps running because it was never depending on a connection.
Straight answers on how this actually works
No. Transcription runs through an open source Whisper model on your own machine. Audio never leaves your device to get turned into text.
Yes. Recording, transcription, and filing all happen locally, so turning off your wifi mid-call doesn't stop any of it working, exactly like the demo above.
The full source is on GitHub under the MIT license. Read the code, run it yourself, or watch the network requests it makes. There's nothing to trust blindly.
Yes. Local here is structural, not a policy. There's no server in the loop to make a promise about in the first place.