FluidVoice v1.6.13-beta.2

What's New

• Edit Mode (Beta) gets its own Edit 1.1 model on Apple Silicon, separate from dictation, with clearer download and failure messages. API providers remain available.
• Fluid 1.1 Splash v2 brings updated local cleanup with first-party model downloads and updates.

Improvements

• AI Providers is now AI Settings, with Edit Mode below the providers. Cleanup Styles keeps Basic, Smart, Default and custom instructions.

• Set up optional Screen Capture beside Read my screen or from Dashboard. Permission is requested only when you click its setup button.

• On Apple Silicon, Pro, Max and Ultra chips with at least 16 GB get a Splash recommendation; other chips keep Pico.

• On Apple Silicon, Fluid Speech Pro, Mini and Pico are available for the existing 25-language list, including compatible Pro selections for meeting transcripts and language redo.

• Meeting transcripts and language redo use your selected compatible Fluid Speech model.

• Use Meeting Home to return from a finished meeting to recording setup.

• Redo a finished meeting transcript in another supported language using its saved audio.

• Choose an Apple Speech language and region in Voice Engine settings on macOS 26 or later.

• The meeting shortcut opens an app and microphone picker first. Press the shortcut again or Return to record; Escape cancels.

• The picker puts your most recently selected meeting app first and suggests the active app when Automatic cannot find a call.

• The meeting picker lists installed browsers and call apps before Other apps, disables closed apps, and lets you refresh the list.

Fixes

• Restoring an older settings backup keeps the correct Edit Mode provider.
• Turning off Read my screen discards captured window context, including after a quick Off/On change.
• Screen-word matching handles everyday contractions without treating them as rare names.
• Prevent TXT and JSON exports from overwriting temporary files or mixing transcripts during concurrent exports.
• Keep the last loaded meeting history visible and show an error when saved meeting files cannot be read.
• Keep the previous meeting transcript and language when a redo fails, is cancelled, or is interrupted by an app restart.
• Google Meet browser detection recognizes the new /meet/<room-code> URLs as well as the original room links.
• Preserve opening microphone speech while meeting echo filtering warms up, without showing a false filtering-failure warning.
• Transcribe very short meeting audio and final snippets without adding words outside the recorded audio.
• More reliable dictation paste in slow editors and apps that handle keyboard shortcuts differently.
• Unread dictation stays on the clipboard after one second for manual recovery, even when Copy to Clipboard is off.
• Paste recovery waits for slow editors, avoids stale or repeated alerts, and better detects failed replacements of selected text.
• Improved text-field discovery in Electron apps and restored dictation in Muse's message box.
• Short imported audio and brief final chunks are no longer skipped.
• Newly recorded shortcuts distinguish left and right Option, Command, Control and Shift. Re-record existing shortcuts to choose a specific side.
• Online meetings no longer silently fall back to microphone-only recording. Choose In person when you only want the microphone.
• The meeting picker keeps the selected microphone when refreshing and asks for a replacement if it disconnects. In person settings preserve the remembered online app.

Contributors

Thanks to @ParamKataria, @TokuiNico and @sabbah13 for meeting language redo, the Apple Speech language picker, short-audio fixes, meeting-history recovery and reliable transcript exports.