Recording safety cap + runtime GPU settings + optional audio cues (fixes #25, #23, #24) - #26
Merged
Merged
Conversation
In hold-to-record mode a recording had no maximum duration: if the global keyboard listener observed a hotkey press but never the matching release, the microphone stayed active indefinitely, and audio could leak from one mode/session into another via the shared Recorder. - Arm the MAX_RECORD_SECONDS safety timer in hold mode too, not just toggle. On timeout in hold mode the mic stops and the (overrun) audio is discarded rather than transcribed and pasted into whatever window has focus. - Track recorder ownership with a single owner + generation counter so a stale timer self-cancels and a stop for a non-owning mode is ignored. - Discard an inherited buffer when a new session starts while the recorder is still owned, preventing cross-mode audio contamination. - Add HotkeyListener.reset_dictation_state/reset_assistant_state so a lost key-release no longer wedges the hotkey (stuck _pressed flag). - Tests: test_recording_safety.py (ownership, hold/toggle timeout, stale timers) and reset-state cases in test_hotkey.py. - Also ignore local release .zip packages. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0197YwE2oqsrH6JmPPRUJ66M
Packaged-exe users could not enable CUDA or change compute precision without editing config.py, even though the transcriber already honours DEVICE and COMPUTE_TYPE. Expose both through the existing settings mechanism, following the same DB-persisted pattern as the other runtime settings. - Settings window: a "Performance" section with a compute-device dropdown (CPU / CUDA) and a compute-precision dropdown (int8 / float16 / float32), a hint about the CUDA runtime requirement, and a restart-required note. - main._load_settings reads and validates the persisted "device" and "compute_type" values (invalid values are ignored, keeping the defaults). - Localised the new labels/hint in EN, IT and DE. - Tests: test_runtime_settings.py covers the callbacks and the load path, including invalid-value rejection. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0197YwE2oqsrH6JmPPRUJ66M
Off by default; enable from Settings to get eyes-free confirmation and a pacing cue so you don't clip the start of your speech. - audio_cues.py: renders two short WAV tones in memory (high on start, low on stop) and plays them with winsound.PlaySound(SND_MEMORY | SND_ASYNC) on a daemon thread. Non-blocking, so recording and hotkey callbacks are never delayed; any playback failure is logged at debug and swallowed. No-op off Windows. - Cues are wired into the dictation path only: start in _on_hotkey_press, stop in _on_hotkey_release, and the hold-mode safety-stop plays the stop cue too — exactly one stop cue per session across hold/toggle/timeout. - Settings: a "Recording sound cues" toggle, persisted as the "audio_cues" setting and loaded in main._load_settings. Localised in EN, IT and DE. - Tests: test_audio_cues.py (valid WAV, SND_MEMORY playback, swallowed backend errors, off-Windows no-op, distinct pitches) plus audio-cue cases in test_runtime_settings.py. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0197YwE2oqsrH6JmPPRUJ66M
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Addresses three open issues reported by @MaxiKingXXL. Split into three self-contained, per-issue commits - reviewable one at a time.
fixes #25- Hold-mode recording safety cap + session isolationIn hold-to-record mode a recording had no maximum duration: a missed key-release left the mic recording indefinitely, and audio could leak across modes via the shared
Recorder.MAX_RECORD_SECONDSsafety timer in hold mode too (was toggle-only). On timeout in hold mode the mic stops and the overrun audio is discarded, never pasted into the focused window.HotkeyListener.reset_dictation_state/reset_assistant_stateso a lost key-release no longer wedges the hotkey.fixes #23- Compute device & precision configurable at runtimePackaged-exe users couldn't enable CUDA or change precision without editing
config.py.main._load_settingsreads and validates the persisteddevice/compute_type(invalid values ignored, defaults kept).fixes #24- Optional audio cues for dictation start/stopOff by default; opt-in from Settings for eyes-free confirmation and pacing.
audio_cues.py: renders two short WAV tones in memory (high on start, low on stop) and plays them withwinsound.PlaySound(SND_MEMORY | SND_ASYNC)on a daemon thread - non-blocking, failures swallowed, no-op off Windows.Testing
test_recording_safety.py,test_audio_cues.py,test_runtime_settings.py, plus reset-state cases intest_hotkey.py.