digimata/quill

Ultra-minimalist macOS recording + transcription.

View on GitHub ↗Jump to charts ↓

Summary Information

Updated 36 minutes ago
Added to GitGenius on September 17th, 2026
Created on July 24th, 2026
Open Issues & Pull Requests: 21 (+0)
GitHub issues: Enabled
Number of forks: 277
Total Stargazers: 3,903 (+0)
Total Subscribers: 14 (+0)

Repository Insights (GitGenius)

Median issue/PR response: 5.6 hours
Mean response time: 14.9 hours
90th percentile: 45.9 hours
Tracked items: 5

Most active contributors

Sign in to see contributor activity.

Charts & Analytics

Fetching additional details & charts...

Issue Activity (beta)

Open issues: 9
New in 7 days: 0
Closed in 7 days: 0
Avg open age: 49 days
Stale 30+ days: 8
Stale 90+ days: 0

Recent activity

Opened in 7 days: 0
Closed in 7 days: 0
Comments in 7 days: 0
Events in 7 days: 0

Top labels

No label distribution available yet.

Most active issues this week

Sign in to see which issues are moving.

Detailed Description

Quill is a minimal macOS meeting recorder and transcriber that runs entirely on your machine.

The tool solves the problem of capturing both sides of remote meetings while keeping all data local. It records your microphone and system audio as two separate tracks, then automatically transcribes both on-device using a speech model. The two-track approach provides natural speaker diarization—distinguishing "me" from "them"—without needing a speaker identification model. Audio is stored in CAF format, which requires no finalization pass, so recordings remain readable even if the process crashes mid-meeting. Transcripts are generated with speaker tags and timestamps, and the tool writes output in both JSON and Markdown formats.

Quill suits developers and professionals who need meeting records but cannot send audio to external services. It requires macOS 15 or later and Apple Silicon is recommended for transcription speed, which runs at roughly twenty seconds per hour of audio. The tool uses Parakeet TDT 0.6B v2 as its default transcription engine via FluidAudio's Core ML port; models download once on first use and are cached locally. A Whisper engine is planned as a fallback option. The interface is deliberately minimal—a single menu-bar icon with one click to start and stop recording—and the entire application is a single Swift binary with no app bundle.

Development shows consistent maintenance with regular commits addressing user-reported issues and feature refinements. The project incorporates user feedback directly into prioritization, responding to specific requests for functionality improvements. Bug fixes are deployed promptly when problems are identified. The maintainer actively documents the tool's behavior and limitations, including clear guidance on system requirements and performance characteristics.