| Version | Date | Notes |
|---|---|---|
1.0.11 |
2026-07-17 | Performance - Prefer ONNX Runtime's CUDA execution provider on Windows x64 so compatible NVIDIA GPUs accelerate local Parakeet transcription. - Keep the CPU provider as a non-fatal fallback when CUDA or cuDNN is unavailable. Windows Runtime - Bundle the ONNX Runtime CUDA and shared provider DLLs in the Windows installer. - Discover the packaged provider directory and versioned cuDNN 9 runtime directories installed under Program Files, including NVIDIA's version-specific bin layout.- Add a release-workflow gate that rejects Windows installers missing either required provider DLL. - Log the requested execution-provider order for direct runtime verification. Upstream Review - Review Murmure 1.10.1 and current main; retain HyperYap's curated-extension approach because the two projects have substantially diverged and the remaining recent fixes depend on Murmure's newer audio architecture. |
1.0.10 |
2026-06-19 | Model - Switch the default speech model to the English-focused NVIDIA Parakeet TDT 0.6B v2 INT8 ONNX package from smcleod/parakeet-tdt-0.6b-v2-int8.- Update installer and CI model download paths to fetch the selected model from Hugging Face. - Add explicit NVIDIA and smcleod model attribution in NOTICE and documentation. Release - Remove BenjaminTerm from the default Windows installer and current documentation; HyperYap now installs as its own voice-to-text app with the optional hotkey helper. Maintenance - Fold in low-risk Dependabot patch/minor updates for frontend, Tauri plugin, and Rust dependencies before release. |
1.0.9 |
2026-04-24 | Release - Update the Windows installer to resolve BenjaminTerm through the latest stable avalonreset/benjaminterm GitHub release.- Confirm the release path selects the current benjaminterm artifact naming used by BenjaminTerm v1.4.3.- Print the selected BenjaminTerm release tag during installation so release verification can confirm the bundled terminal version path. |
1.0.8 |
2026-04-24 | Features - Add the first HyperYap terminal editing layer in the Windows hotkey daemon. - In supported terminals, Ctrl+C now attempts terminal copy through Ctrl+Shift+C before falling back to Ctrl+C interrupt behavior when no selected text is copied. - Track recent HyperYap-managed terminal pastes and let Ctrl+Z erase the last pending paste with bounded backspaces. Release - Prepare README, installer docs, and GitHub release workflow for Windows, Linux, macOS Apple Silicon, and macOS Intel artifacts. |
1.0.7 |
2026-04-22 | Bug Fixes - Move smart paste clipboard waits and image conversion off the low-level keyboard hook owner thread so terminal Ctrl+V interception stays responsive under fast repeated screenshot paste attempts. - Add an in-flight guard for smart paste so repeated Ctrl+V presses do not leak raw V characters while a screenshot paste is resolving.- Only skip HyperYap's own recent synthetic key events, allowing user remappers and macro tools to still trigger smart paste. |
1.0.6 |
2026-04-21 | Performance - Add a native prepared screenshot pipeline in the hotkey daemon so Win+Shift+S and PrintScreen captures are saved as PNG paths before terminal Ctrl+V. - Replace the normal PowerShell image conversion path with native Rust clipboard DIB handling, PNG encoding, clipboard text swap, and image restore. - Keep PowerShell image handling as a fallback only. Bug Fixes - Make speech-to-text paste terminal-aware on Windows so vanilla WezTerm receives Ctrl+Shift+V text paste instead of Codex image paste handling. |
1.0.5 |
2026-04-21 | Bug Fixes - Make smart screenshot paste work in vanilla WezTerm by using the terminal paste chord after HyperYap converts images to saved PNG paths. - Track recent Windows screenshot shortcuts so fast Ctrl+V waits for the screenshot image instead of falling back to stale text paste. - Update the bundled AutoHotkey fallback script with the same WezTerm-safe paste behavior. |
1.0.4 |
2026-04-21 | Bug Fixes - Add a 750ms safety window to the tracked Ctrl state used by smart paste so a missed Ctrl keyup cannot cause later plain v keystrokes to be intercepted.- Keep the fast Ctrl+V race fix while reducing the risk of stale modifier state side effects. |
1.0.3 |
2026-04-21 | Bug Fixes - Fix a fast Ctrl+V race where Windows could report Ctrl as not held and let a raw v through to BenjaminTerm before smart paste handled the screenshot.- Track physical Ctrl state inside the low-level keyboard hook and suppress the matching V keyup even if Ctrl is released during smart paste handling. |
1.0.2 |
2026-04-21 | Bug Fixes - Prevent terminal smart paste from typing a stray v when Windows has not finished publishing a screenshot to the clipboard.- Keep terminal text paste fast while waiting longer for empty or unstable screenshot clipboard states. - Extend the empty clipboard screenshot wait window to 5 seconds before suppressing paste. |
1.0.1 |
2026-04-21 | Bug Fixes - Harden smart screenshot paste for terminals by waiting for Windows clipboard image formats before falling back to normal paste. - Run clipboard image conversion in STA PowerShell with retry handling so Snipping Tool screenshots are captured reliably. - Restore the exact pasted screenshot image after paste instead of restoring the latest file from the screenshots folder. - Update the legacy AutoHotkey smart paste scripts with the same image format retry behavior. |
1.0.0 |
2026-04-04 | Forked from MURmure 1.8.1 and rebranded as HyperYap. Bundled Parakeet model, preset configs, AutoHotkey scripts, and PowerShell installer for grab-and-go local speech-to-text on Windows. |
The entries below are the original MURmure changelog preserved for reference. They document the upstream project's history prior to the HyperYap fork.
| Version | Date | Notes |
|---|---|---|
1.8.1 |
2026-04-04 | 🐛 Bug Fixes - (macOS) Fix shortcuts not working - Space and other shortcuts were not detected on macOS. Replaced rdev event listener with polling-based approach. - Remove Ctrl+B shortcut that could hide sidebar (#266). - Allow virtual microphones in device list, e.g. Nvidia Broadcast (#164). - Add beta tag and CPU usage warning for Voice Mode. - Security updates for dependencies. |
1.8.0 |
2026-03-28 | ✨ New Features - (Major) Voice Mode - Activate Murmure by voice with customizable wake words, no keyboard needed. Includes auto-send enter, and wake word listener stays active during keyboard-triggered recording. - (Major) LLM Remote Servers - Connect to remote OpenAI-compatible API servers for LLM Connect, with per-mode provider switching, model search, per-mode voice triggers, and onboarding flow. - (Major) Import/Export Settings - Import and export configuration settings partially or fully. Add CLI import support ( murmure import <file>) for sysadmin mass deployment.- Shortcuts - Add support for mouse buttons, OEM keys (-, =, [, ], ;, ', etc.), <>, ², F13-F24. Add a shortcut in overlay to cancel recording. - Rules - Add regex support for custom rules, improve rule labels, add "?" helper for replacement text, short text correction (auto-lowercase/remove punctuation), and rule reordering. - LLM - Allow LLM mode reordering. Add "Voice Dictation" preset for interpreting spoken formatting commands as typographic characters. - Dictionary - Add option to clear all dictionary entries. - About Page - Redesigned with starry background and sponsor cards. - Autostart - Start minimized to tray on autostart. - macOS - Add configurable option to show/hide application in Dock. - Linux - Add .deb package and install.sh one-liner script for Debian/Ubuntu.- Update - Update tauri-plugin-updater to 2.10.0 for multi-format support (.deb, .exe, macOS Intel).- Sidebar - Add "Release notes" link in sidebar footer. - Defaults - Improve default settings for new users (transcription in RAM, trailing space). 🐛 Bug Fixes - Fix overlay scrollbar on Windows, overlay stealing focus, overlay freeze on audio timeout. - Fix double percentage display during download. - Use dark mode colors consistently. - Improve microphone list and keep manually selected device when disconnected. Fix crash when no microphone is available. - Improve LLM accuracy with system prompts. Improve response speed of reasoning models by disabling thinking mode. - Improve smart mode auto-spacing and word deduplication (parakeet fix). - Fix selected text detection when clipboard already contains the same content. - Fix ToggleToTalk accidental restart caused by X11 auto-repeat. - Fix spurious modifier key release on macOS after paste and recording lock management. - Fix wake word language-aware validation, history update, and EMA smoothing for VAD. |
1.7.0 |
2026-02-12 | ✨ New Features - (Major) Select Command - Select text and modify it with a custom command (e.g. fix grammar, translate to English, etc.). - (Major) LLM Prompts - Support for multiple saved prompts with syntax highlighting for the prompt editor, anchor tags enforcement, and a new "Cursor Developer" preset. - Input Microphone Selection - Allow selecting the input microphone in Settings. Thanks to @litel-fr. - Dictionary Import/Export - Import and export words from the dictionary, including medical presets. Thanks to @icristescu (#72). - Text Insertion Mode - New setting to choose between standard, terminal-compatible, or direct typing insertion modes (#119). - Log Verbosity - Configurable log verbosity levels (trace, debug, info, warn, error) and a button to open the logs folder. 🐛 Bug Fixes - Fixed full screen issue on Select Model page on macOS (#82). - Transcription is no longer part of the logs by default (#88). - Temporary audio is now saved in the tmp folder instead of app_dir in case of crash (#88). - Fixed invalid word format error message in dictionary. - Allow digit conversion threshold to be 0. - Updated dependencies to fix security vulnerabilities. ♻️ Refactor - Fixed main SonarQube issues. - Streamlined shortcuts logic and optimized event lookup. 🔧 CI - Added Security Scanning with SonarQube. - Improved static analysis for contributions (test, compilation, linting, etc.). - NSIS exe installer now runs without administration privileges. |
1.6.0 |
2025-12-15 | ✨ New Features - (Major) LLM Connect - Connect Murmure to a local LLM (Ollama) to enhance, correct, or reformat transcriptions through post-processing. Includes presets for translation, medical dictation, development, and automatic reformulation. - (Major) Formatting Rules - Automatic text transformations: add trailing spaces, insert spaces before ?!, convert text-to-numbers, or apply custom find/replace rules.- Toggle-to-Talk Mode - New alternative to push-to-talk: press once to start recording, press again to stop. Thanks to @nicolascrop. - Sound Feedback - Audio cues when starting and stopping a recording. ♻️ Refactor - Improved UI notifications. - Backend refactor aligned with SRP principles. - Dictionary configuration migration in preparation for import/export support. Thanks to @icristescu. 🐛 Bug Fixes - Fixed overlay misalignment with fractional display scaling. Thanks to @flomillot (PR #73). |
1.5.0 |
2025-11-15 | ✨ New Features - Display words per minute, MB saved (no cloud), and total words statistics - MacOS version support - Option to not make the last transcription persistent in history and remove it at each shutdown - FR/EN translation of the application - Add onboarding tutorial for first time users 🐛 Bug Fixes - Prevent menu from scrolling on long pages - Restart application after uploading to new version - Display the overlay on the active screen |
1.4.0 |
2025-11-01 | ✨ New Features - Click on history, save it in clipboard - Add a button to clear transcription history - Add option in Settings to keep the transcription in clipboard 🐛 Bug Fixes - Second transcription losing cursor focus and sometimes not working properly. - Minor UI change on long shortcut in Settings |
1.3.0 |
2025-10-25 | ✨ New Features - Added an experimental API allowing Murmure to connect with external software (can be enabled in Settings → System) - Improved shortcut UI for better usability - Removed the experimental tag for the Linux AppImage. It now works identically to the Windows version (users who had the experimental version need to reinstall it to benefit from the "check for updates" feature) 🐛 Bug Fixes - Fixed small memory leak - Fixed the unwanted "v" key activation when assigning shortcuts - Corrected the GitHub link |
1.2.1 |
2025-10-17 | Fix overlay position + check for updates button + signed msi + Linux experimental AppImage |
1.2.0-alpha |
2025-10-14 | Add Overlay |
1.1.0-alpha |
2025-10-13 | Add 'Past last transcript' shortcut |
1.0.0-alpha |
2025-10-13 | Initial version |