feat: add --listen-during-wake-sound argument#273
Conversation
|
oh perfect i needed this |
|
Cool feature <3 |
Removed TIMER_MAX_RING_SECONDS from configuration options.
|
Seems good, but make sure that this won't cause confusion with STT. |
It will depend heavily on your setup (e.g. distance between mic and speaker, AEC enabled, etc), your wake sound (e.g. beep vs "yes, sir"), and your STT model (e.g. if it will transcribe "beep" literally). Anyway, the feature is disabled by default. |
|
Id add in the docs that this feature does heavily depend on effective AEC, and mention the pipewire software aec section as well as hardware stuff |
Can you check the latest version of the PR? I tried to add a link to the existing AEC doc file. I'm not sure if I agree that it depends heavily on the AEC, though, since the default wake sound is just a beep, which the STT should in the worst case treat it like an onomatopoeia, adding literally a "beep" to the phrase. |
|
Less about onamonapea and more that bad aec + bad stt would struggle to pick up on voice over the audio But also sorry I'm very drunk right now 😅 |
Add listen_during_wake_sound field to ServerState (default true).
Add listen_during_wake_sound option to AppConfig (default true, preserving fork's existing behavior). When false, LVA waits for the wakeup sound to finish before streaming audio to HA, useful for setups without effective AEC.
Split _start_conversation into two paths based on listen_during_wake_sound. When true (default), streams audio immediately with fire-and-forget wakeup sound (existing behavior). When false, plays wakeup sound with done_callback to _on_wakeup_sound_finished, which starts streaming after the sound completes. Port upstream PR OHF-Voice#271: Promote voice event logging from DEBUG to INFO in handle_voice_event for pipeline visibility without --debug.
Pass config.app.listen_during_wake_sound to ServerState constructor.
Document listen_during_wake_sound option with AEC guidance.
Add listen_during_wake_sound field to ServerState (default true).
Add listen_during_wake_sound field to ServerState (default true).
|
Hm, we are missing the changes in the .env.example file. If we could add that shortly this can be added in the planned release. |
|
@florian-asche Done. |
|
testing this now... |
Remove unused TIMER_MAX_RING_SECONDS argument handling
From my comment at #170 (comment)
I also added a small fix for the doc of other command line argument