v1.1.0 — Archival Model Choice

Call Assistant captures both sides of a meeting without joining it as a bot, transcribes locally on your Mac, and runs a Claude sidecar live during the call. Recordings, transcripts and notes stay in your Library folder.

This is the first public release.

Install

Download CallAssistant_1.1.0_aarch64.dmg, open it, and drag the app to Applications. The app is signed with a Developer ID and notarized by Apple, so it opens with the standard first-run confirmation and no security bypass.

The guide covers first-run setup: the audio driver, the speech model download, and where to put a Claude API key.

Requirements

• macOS 14.4 or later, Apple Silicon
• ~1.5 GB of disk for the default speech model, downloaded in the app
• A Claude API key for the AI features. Capture and transcription work without one, and with the AI features off the app makes no API calls.

What is in this release

The archival re-pass is no longer Whisper-only. You choose which downloaded model runs it, from a tiered picker: large-v3 is the one Supported choice, and everything else — including the new NVIDIA Parakeet engine — is marked Experimental until it earns a promotion on real call audio. A hand-install slot takes models the app does not download itself.

Speech engines are now plugins, so adding one cannot bloat or destabilise the app binary. Parakeet runs as a separate helper process rather than being linked in.

Existing installs migrate their model storage on first launch. Nothing is deleted and nothing is re-downloaded.

Verifying the download

SHA256SUMS lists both artifacts. Put it beside the downloaded file and run:

shasum -c --ignore-missing SHA256SUMS

The .tar.xz is what Sparkle downloads for automatic updates; you do not need to fetch it by hand, which is what --ignore-missing is for.

Links

Guide · Privacy · Licence · Third-party notices · Support