Talkink is designed to keep your data on your machine:

• Speech & transcripts never leave your Mac. Transcription runs locally on Apple Silicon via MLX. Audio is processed in memory; transcripts are stored only in a local file (~/Library/Application Support/Soyle/history.json).
• No telemetry, no analytics, no accounts.
• Network calls are limited to two, both inspectable in the source:
  1. Model downloads (~2.5 GB for the default model; 760 MB–4.1 GB depending on your pick) from Hugging Face into ~/.cache/huggingface (Sources/SoyleKit/ModelDownloader.swift).
  2. An optional Sparkle update check against the EdDSA-signed appcast at talkink.app (Sources/Soyle/Updater.swift) — disable it in Settings. Update archives download from GitHub Releases.

• Microphone — to record while the push-to-talk key is held.
• Input Monitoring — to detect the push-to-talk key globally (listen-only event tap).
• Accessibility — optional; only used to paste at the cursor (synthetic ⌘V). Without it, Talkink falls back to copying to the clipboard.

Please open a GitHub issue, or for sensitive reports contact the maintainer privately via the email on the GitHub profile. We'll acknowledge and respond as quickly as we can.