Changelog
Release Notes
A full history of every V-Ghost release — features, fixes, and improvements.
- Gamepad Macros — Voice-control a virtual Xbox 360 controller via the bundled ViGEm Bus driver (no separate install needed):
- Gamepad Button Press — Press and immediately release any button (A, B, X, Y, LB, RB, LT, RT, Start, Back, D-Pad directions).
- Gamepad Button Hold — Hold a button for a configurable duration (ms) before releasing. Button release is guaranteed even if a cancellation token fires.
- Gamepad Button Chord — Hold a modifier button, then press a second button while it is held (e.g. LB+A), with configurable pre-hold, press, and post-hold timing.
- Gamepad Axis — Set an analog stick or trigger to a precise value (0–100) for a configurable duration. Useful for throttle, steering wheel, or look-axis control.
- Key Hold Macro — Hold a keyboard key or modifier combination for an exact number of milliseconds, then release. Game engine registers it as a held input rather than a tap — great for thrust/boost inputs or melee charges.
- Key Repeat Macro — Press a key N times (1–1 000) with an optional inter-press delay. Replaces manual Repeat + Key Press sequences with a single compact action.
- Type Text Macro — Simulate character-by-character typing into the focused window at a configurable speed. Supports
{variable}interpolation — say a command to type a stored phrase. - Gamepad Push-to-Talk — Any physical XInput controller button (A/B/X/Y/LB/RB/LT/RT/Start/Back/D-Pad) can now be assigned as the push-to-talk key alongside keyboard keys in Settings.
- Free Trial — V-Ghost can now be downloaded and evaluated at no cost. A license key (available from the pricing page) unlocks full functionality.
- Whisper Recognition Engine — Ultra-lightweight offline speech recognition using OpenAI Whisper models (tiny quantized GGML). Lower latency than Vosk via speculative rolling inference: results fire before the speaker finishes. Single-pass setup; language auto-detected.
- Persistent Variables — Variables now survive app restart. Stored per-profile on disk in JSON format; shared across all commands in that profile regardless of when they run. Perfect for tracking state like game modes, settings, or counters.
- Dictation Macro — Capture free-form spoken text and either type it into the focused window or store it in a named variable. Useful for note-taking, chat input, search queries, or any ad-hoc text entry.
- HTTP Request Macro — Send GET or POST requests with
{variable}interpolation in URL and body. Capture response headers/body into a variable for conditional logic. Enables voice-triggered webhooks, API calls, and integrations. - Repeat Macro — Execute a sub-action list a fixed number of times (1–1000) with optional delay between iterations. Use cases: spam-checking (repeated key presses), multi-item selections, or choreographed sequences.
- Repeat While Macro — Loop a sub-action list until a variable condition becomes false. Uses the same 8 operators as If–Then–Else. Includes iteration guard (max 100 by default to prevent hangs) and optional inter-iteration delay.
- Audio Device Management — Real-time monitoring of microphone and speaker connections. Gracefully handles device hot-plugging; recognition engine auto-recovers when audio input is unplugged and replugged.
- Improved UI & Action Editor — New About page showing app version and credits. Action editor refactored with ActionTypeRegistry for maintainability. Command editor expanded with better sub-action nesting UI (Repeat and RepeatWhile bodies can be edited inline).
- All-new action configuration mapper with full JSON roundtrip validation and 482 test cases to prevent data loss.
- Zero breaking changes — all v1.3.0 profiles import and run unchanged.
- Variables & State Management — Store, toggle, and increment named values that persist across commands within a profile session. Use
{VarName}syntax anywhere to interpolate values into text. - If–Then–Else Conditionals — Branch command execution based on variable state. Supports 8 operators:
Equals,NotEquals,IsTrue,IsFalse,IsSet,IsNotSet,GreaterThan,LessThan. Branches can be nested to any depth. - Call Command — Invoke another command in the profile as a subroutine, sharing the same variable context. Recursion protection prevents infinite loops (max depth 10).
- Text-to-Speech (Say Text) — Speak dynamic text via Windows SAPI with configurable rate (−10 to +10), volume (0–100 %), and voice selection. Supports
{variable}interpolation. - Audio Playback (Play Sound) — Trigger WAV, MP3, FLAC, OGG, or WMA files at a configurable volume. Playback completes before the next action runs.
- Mouse Move — Move the cursor to an absolute screen position or by a relative pixel offset.
- Mouse Scroll — Scroll vertically or horizontally by a configurable number of clicks.
- Command Groups — Assign commands to named groups for collapsible sections in the profile editor. No effect on recognition or execution.
- Global Profiles — Mark a profile as global so its commands are always active alongside whatever game profile is running.
- VR Overlay — A floating in-headset overlay displays the active profile name and the last recognised command in real time. Works with OpenVR / SteamVR headsets; no headset removal required to see what V-Ghost heard.
- First-Run Onboarding Wizard — Guided setup for microphone selection, recognition engine, language model download, and first profile creation.
- Built-in profile templates — import pre-made profiles for common applications
- Command aliases — set multiple phrases that trigger the same command
- Improved execution log — export command history to CSV for analysis
- Light theme support — toggle between dark and light UI
- Settings persistence — auto-save theme, window position, and application state
- Onboarding flow — guided first-run experience for new users
- Bug fix: Confidence scores not persisting across app restarts in some cases
- Bug fix: Fuzzy matching not working for commands with special characters
- Bug fix: Acronym expansion causing false positives (e.g., "FSD" now correctly expands)
- Improved error messages when recognition engine fails to initialize
- Performance: Reduced memory usage during long listening sessions
- Bug fix: Vosk model download failing on some network configurations
- Bug fix: Execution log not showing results for timed-out commands
- Bug fix: Audio input not initializing when multiple microphones detected
- Improved startup time when loading large profiles (100+ commands)
- Execution log viewer — detailed table showing all command executions with timestamp, confidence, and result
- Fuzzy matching — Levenshtein distance algorithm now catches minor speech recognition errors
- Acronym expansion — spell out acronyms phonetically (e.g., "FSD" recognized from "eff ess dee")
- Alternative recognition results — show multiple hypotheses when Vosk returns alternatives
- Drag-and-drop macro actions — reorder command steps in the editor by dragging
- Command execution statistics — track success rate and average execution time per command
- Window position persistence — app remembers last position and size on launch
- Offline speech recognition via Vosk — no internet required after model download
- Windows SAPI fallback engine — use built-in Windows speech recognition
- Keyboard macro actions — simulate any key press or modifier combination
- Mouse macro actions — left-click, right-click, middle-click, move to coordinates
- Process launcher — open any application or run any script by voice
- Delay action — configurable pause between steps in a macro sequence
- Macro sequences — chain multiple actions into a single voice command
- Profile management — unlimited named profiles with per-profile language and engine
- Auto-activate profiles — automatic switching based on the foreground process
- Push-to-talk mode — optional hotkey to gate recognition
- On-screen overlay — shows active profile and last recognized phrase
- System-tray minimize — runs quietly in the background
- Language support: 30+ languages including English, German, French, Spanish, Russian, Chinese, Japanese, and more
- Dark VS Code-inspired theme throughout
- Confidence threshold control per profile