OpenWhisperer 2.0.8

Small fixes that matter day to day, and the app now tells you when a newer version is out.

• Hold-to-talk no longer sticks in recording. Pressing a shortcut while holding the trigger key — Ctrl+C with Ctrl as the trigger, which terminal users do all day — used to leave the recording running until your next tap, which then typed everything captured in between. A shortcut during a hold now cancels that recording. A key release macOS never reports (password fields, Secure Keyboard Entry terminals) is also recovered within about half a second. (#26)
• Mute spoken replies from the menubar. A Mute Spoken Replies toggle silences every agent's spoken replies — for a call, or music — without touching dictation. Muting stops whatever is playing; the menubar icon shows a slashed speaker while it's on, and the agent is told its reply wasn't spoken, so it won't retry. (#51)
• The app tells you when a newer version is out. Once a day it asks GitHub for the latest version number — nothing about you or what you say is sent — and offers the release page in the menubar and in Settings → General. It never downloads or installs anything. Turn it off in Settings → General if you'd rather it didn't ask.
• Changing the TTS port no longer disconnects your agents. The port in Settings → Advanced wasn't saved, and every agent was pointed at port 8000 regardless, so changing it quietly took the speak tool away from all of them. The port now sticks once the server has actually bound it, and connected agents follow it. A port another process already holds is reported as an error and changes nothing.
• Long dictations while translating no longer time out. The transcription watchdog allowed a flat 35 seconds, which the slower translating model ran past somewhere beyond a minute of speech — losing the whole recording. The budget now grows with the length of what you said. Recordings also stop themselves at ten minutes, so a forgotten press-to-talk can't record forever; the transcript is kept in Record Voice & Text rather than typed into whatever app happens to be in front.
• The README's Install link pointed at the previous version. Fixed, and now checked at build time. There's also a proper first-launch guide for a new Mac: Gatekeeper on macOS 15, and why Accessibility and Microphone must be granted again after every update.
• Speech engine updated to WhisperKit 1.1.1 (crash fixes; same speed).

Install: download the DMG below, drag to Applications. On macOS 15 and later, open it once, dismiss the warning, then System Settings → Privacy & Security → Open Anyway. Or from Terminal, before opening:
xattr -d com.apple.quarantine ~/Downloads/OpenWhisperer-2.0.8.dmg

Updating from 2.0.7: grant Accessibility and Microphone again (remove and re-add OpenWhisperer under System Settings → Privacy & Security). The build isn't Developer ID signed, so macOS treats every release as a new app.