Windows and Linux early access status

Recordist is made for the Mac first. The Windows and Linux versions exist, are produced from the same code, and are labelled early access on the download page for one honest reason: as of 22 September 2026 we have not yet verified a full recording on either platform ourselves. Recording may fail, or capture only one side of a call.

What is implemented

• Windows: system audio through WASAPI loopback (no virtual driver), microphone through the standard capture API, Whisper with CUDA or Vulkan where a GPU is present and CPU otherwise. Installer is an .exe.
• Linux: system audio through the PipeWire or PulseAudio monitor source, X11 and Wayland, Whisper with Vulkan or CPU. AppImage and .deb.
• Everything above the capture layer (database, transcription, notes, search, MCP, A2A, exports) is the same code as on macOS and is covered by the same tests.

What is not verified

Whether the capture backends actually produce clean two-channel audio on real hardware, whether detection names apps correctly, and whether the tray icon and prompt behave. These are exactly the things that need a person on a real call.

Signing

Windows builds are not yet signed, so SmartScreen will warn. Verify the checksum from SHA256SUMS.txt before running the installer.

How to help

Try a call. Then open a ticket with the platform, the app version, whether both sides of the call were transcribed, and Settings → General → Export diagnostics (no audio, no transcripts). One confirmed call on each platform is what moves them out of early access, and this page and the download page will say so when that happens.