Fully offline · works with Wi-Fi off

Recording & transcription
that never leaves your machine.

Wisp captures your microphone and the other side of the call at the same time, then transcribes both on-device. No cloud APIs, no uploads — just a clean, local transcript.

Download for macOS View source

Apple Silicon · macOS 26 (Tahoe) · Windows & Linux coming soon

Everything stays on-device

A small, focused app that does one thing well — and keeps it private.

Fully offline

Audio and transcripts stay on your device. Turn off Wi-Fi and Wisp keeps working exactly the same.

On-device transcription

Mic + system audio

Core Audio Process Taps capture meeting-app output without prompts, mix it with your mic, and merge both sides into one transcript.

Built in Rust, GPU-rendered

The UI runs on GPUI, the framework behind the Zed editor — native-feeling and smooth.

Simple local storage

Recordings save as WAV, metadata as SQLite. Browse past sessions in the library — easy to export and analyze later.

Yours, and only yours

No accounts, no telemetry, no servers. Nothing to sign up for and nothing leaving your laptop.

How it works

Two audio streams in, one tidy transcript out — every step on your Mac.

1

Capture both sides

A Core Audio Process Tap grabs system audio while your microphone is recorded in parallel — no virtual cables or prompts.
2

Transcribe on-device

Each stream flows through Apple's SpeechAnalyzer, producing live segments that refine themselves as you speak.
3

Merge & store locally

Mic and system segments merge into a single timeline, then persist as WAV + SQLite under your Application Support folder.

 Core Audio Process Tap ─┐
                         ├─► WispAudioKit ─► GPUI desktop
 Microphone input ───────┘        │              ▲
                                  └─► SpeechAnalyzer
                                          │
                                          └─► WAV + SQLite (local)

Privacy by architecture

The network is never in the loop

Wisp doesn't have a cloud backend to leak to. Capture, transcription, and storage all happen locally, so your conversations stay between you and your machine — by design, not by policy.

0bytes uploaded

100%on-device

WAV + SQLiteopen, local formats

Get Wisp

Open-source and free. Built for the latest Apple Silicon Macs.

Download for macOS

macOS 26 (Tahoe) — uses SpeechAnalyzer & Core Audio Process Taps
Apple Silicon (aarch64) build, shipped as a signed .dmg
Microphone & system-audio permission, prompted on first launch

Download latest release Build from source

On the roadmap: Windows (WASAPI loopback) and Linux (PipeWire) support, plus export to Markdown / SRT / JSON.

Recording & transcription that never leaves your machine.