You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This commit was created on GitHub.com and signed with GitHub’s verified signature.
FluidVoice v1.6.2
What's New
Custom Dictionary can now learn corrections from your voice: say a difficult word a few times, then save the misheard versions as one replacement.
Fluid Intelligence can now work with much longer text: up to about 2,000 words instead of about 200, so longer outputs are less likely to get cut off.
Added voice-friendly /commands and app-aware @mentions for Slack and Discord, with autocomplete-friendly spacing.
Added direct punctuation-to-symbol conversion for spoken punctuation. Try phrases like dash, hyphen, question mark, period, comma, slash, and more. It works without Fluid Intelligence or any AI.
Added a Help > Change Logs menu item so you can open release notes from inside FluidVoice.
Improvements
Vocabulary Boosting now explains that it can make transcription slower, so it is easier to decide when to use it.
Fluid Intelligence settings are easier to understand, with clearer speed controls, safer long-dictation guidance, and simpler reset/delete model actions.
Fluid Intelligence can enable a speed boost for up to 15% faster output, using about 100 MB more memory.
Faster recording start is now on by default, so FluidVoice starts listening sooner and is less likely to miss your first word.
Parakeet v2/v3 transcription is faster, and the overlay closes faster after dictation finishes.
History copy buttons now look more consistent.
Auto-converted punctuation is smarter around everyday text, keeping normal phrases, commas near symbols, and slash commands intact while still formatting URL/path-like slashes.
(@) At sign conversion is limited to coding, terminal, AI, and team-chat apps such as Codex, ChatGPT, Claude, Cursor, Windsurf, Xcode, VS Code, Terminal, iTerm, Warp, Ghostty, Slack, Discord, and Teams.
Model setup now shows clearer preparation states so downloads do not look stuck while FluidVoice gets the voice engine ready.
Fixes
Fixed right-side modifier shortcut combinations such as Right Option + Right Shift for primary dictation.
App updates no longer send existing users back through onboarding, while first-time installs still show onboarding normally.