* init attempt at new kb
* 0.1.1 handy-keys
* format
* cleanup
* support for capslock/shiftlock?
* move kb to experimental
* use hook
* fix macos bug
* attempt mouse buttons as well
* feat: add automatic filler word removal from transcriptions
Add filter_transcription_output() that removes filler words (uh, um, hmm,
etc.) and hallucination patterns ([AUDIO], (pause), <tag>...</tag>) from
transcriptions. Inspired by VoiceInk's approach.
- Add regex-based filter in audio_toolkit/text.rs
- Integrate into transcription pipeline after custom word correction
- Add comprehensive tests for filler word removal
- Add regex crate dependency
* feat: collapse repeated 1-2 letter stutters in transcriptions
Add collapse_stutters() to reduce model hallucination artifacts like
"wh wh wh wh wh why" -> "wh why" or "I I I I think" -> "I think".
Collapses any 1-2 letter word that repeats 3+ times consecutively
to a single instance (case-insensitive matching).
* remove bracket/xml stuff
* format
---------
Co-authored-by: CJ Pais <cj@cjpais.com>
* feat: add Moonshine Base speech recognition model
Add support for the Moonshine Base ASR model, which offers:
- 5x faster transcription than Whisper
- Better accent recognition
- ~240MB model size (encoder + decoder + tokenizer)
Changes:
- Update transcribe-rs to v0.2 with moonshine feature
- Add EngineType::Moonshine variant
- Add Moonshine Base model definition (400MB, 80% accuracy, 90% speed)
- Add model loading and transcription logic
- Add i18n translation for Moonshine Base
Note: Model files must be hosted at blob.handy.computer/moonshine-base.tar.gz
Expected structure: encoder_model.onnx, decoder_model_merged.onnx, tokenizer.json
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* Update moonshine parameters.
* add translations
---------
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Co-authored-by: CJ Pais <cj@cjpais.com>
* fix: replace async-openai library with post request to support custom providers
* move models call to llm_client too
* tweak post processing ui and add groq and cerebras
---------
Co-authored-by: CJ Pais <cj@cjpais.com>
* fix: re-assert always-on-top for overlay on Windows
On Windows, the recording/transcribing overlay can lose its topmost
Z-order and fall behind other windows during certain window manager
events (e.g., dragging windows, alt-tabbing, or other apps asserting
focus).
This fix uses the Win32 SetWindowPos API directly with HWND_TOPMOST
to aggressively re-assert the overlay's Z-order immediately after
showing it. Key implementation details:
- Uses Tauri 2's WebviewWindow::hwnd() to get the native handle
- Calls SetWindowPos on the main UI thread via run_on_main_thread
- Applies flags: SWP_NOMOVE | SWP_NOSIZE | SWP_NOACTIVATE | SWP_SHOWWINDOW
- SWP_NOACTIVATE prevents stealing focus from the active window
- SWP_SHOWWINDOW reinforces the topmost state during the show
The fix is called after .show() in both show_recording_overlay and
show_transcribing_overlay, which is more reliable than setting it
before the window is visible.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* keep the fix to windows only
---------
Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: CJ Pais <cj@cjpais.com>
* initial specta commands
* refactor to use the bindings in the ui
* merge fixes
* remove unused commands
* add clamshell to lib
* Remove Invoke from everything.
* fix settings not being loaded properly
* fix settings being overwritten due to deserialization failure + remove
provider kind
* fix model loading text bug
* feat: add conversion between Simplified and Traditional Chinese. If the selected language is Simplified or Traditional Chinese, it will convert the text to the selected variant(Simplified or Traditional) after been recognized, depends on opencc, if it is not installed, the conversion will skip.
* Adopting ferrous-opencc makes cross-platform packaging easier.
* Fix unstable mute implementation on Windows/Linux
* add some delay before the mute so the audio feedback plays
---------
Co-authored-by: CJ Pais <cj@cjpais.com>
* Add LLM post-processing feature with settings management and UI integration
* Refactor LLM post-processing settings and prompt selection logic
* Rename CLAUDE.md to more standard AGENTS.md
* Replace Open Router with OpenAI Compatible endpoints for post-processing
* Fix post processing settings api styles
* refactor or something
* simplify
* cleanup a bit?
* remove useless code
* minor improvements
* fix laggy textarea
* Store post-processing results in db
* fix text color
* new default prompt
---------
Co-authored-by: CJ Pais <cj@cjpais.com>
* Use an optimal release profile... load managers & hotkeys first, add a
splash, simplify app handler
* removed splashscreen, hopefully app boots fast enough anyhow
* No need to load transcription model on start now that its loaded on
demand & cleanup
* Make an unloaded model status
* tiny modification to unloading
* update comment
---------
Co-authored-by: CJ Pais <cj@cjpais.com>
* fix: restore translate-to-english functionality
The translate option broke when migrating from whisper-rs to transcribe-rs.
This fix patches transcribe-rs to add translate parameter support and passes
the translate_to_english setting to the Whisper engine.
Additionally fixed AccessibilityPermissions API calls to match updated library.
Translation only works with Whisper multilingual models (not .en models).
Parakeet models don't support translation as they are transcription-only.
Fixes#152
* bump transcribe rs
---------
Co-authored-by: CJ Pais <cj@cjpais.com>
* wip: model manager support for parakeet
* working parakeet??git status
* Update LanguageSelector.tsx
* Use the published version of the crate.
* move word correction into audio toolkit
