Changelog

WhisperLocal is now WhisperPal. Your library, settings, models and integrations carry over automatically.

New

• Live transcript window — a floating, pinnable transcript that follows the meeting as it happens.
• LAN transcription — pair another computer running WhisperPal and send recordings to it over an encrypted connection. Speaker recognition can run there too.
• Remote MCP for every computer — one connection from Claude or ChatGPT reaches all of your paired computers. Transcript snapshots can optionally answer while a Mac is offline; audio never leaves your device.
• More local engines — WhisperKit, Qwen3-ASR and Parakeet join Whisper, with recommendations for your Mac.
• Bring your own AI — configure several AI and transcription providers, including the Codex, Claude Code and OpenCode CLIs.
• Browser meetings — calls in Chrome, Edge, Safari, Arc and Dia are detected alongside native meeting apps.
• Dictation styles — per-app styles, double-tap modifier shortcuts and push-to-talk.

Improved

• Speaker recognition filters microphone bleed and remembers your corrections.
• Meeting detection, calendar titles and recording reminders are more reliable.
• Settings were rebuilt with clearer navigation.
• AI summaries and chat render Markdown and stream faster.
• Faster library loading and lower memory use for local models.

The first packaged build of the app, then called WhisperLocal.

Highlights

• Delete a meeting’s saved audio while keeping the meeting, transcript, summary and conversation.
• Recording reminders in compact, medium and large sizes.
• Dictation with double-tap left or right modifier keys, including hold-to-talk.
• Self-hosted Cloudflare relay with device approval and OAuth.
• Initial Windows and Linux builds for import, transcription and AI.