Current behavior: If recording is toggled on and immediately off again (e.g. a misclick), OpenFlow still processes and outputs text — typically phrases like "Thank you" or "Thanks for watching" — despite no audio being recorded.
Expected behavior: If the recording window is too short or contains no meaningful audio input, OpenFlow should suppress output entirely and not inject anything into the target app.
Notes: A simple minimum-duration threshold or silence/energy detection before passing audio to the Whisper model would catch this. The output in these cases is clearly Whisper hallucinating from near-silence.
Current behavior: If recording is toggled on and immediately off again (e.g. a misclick), OpenFlow still processes and outputs text — typically phrases like "Thank you" or "Thanks for watching" — despite no audio being recorded.
Expected behavior: If the recording window is too short or contains no meaningful audio input, OpenFlow should suppress output entirely and not inject anything into the target app.
Notes: A simple minimum-duration threshold or silence/energy detection before passing audio to the Whisper model would catch this. The output in these cases is clearly Whisper hallucinating from near-silence.