When Naomi asks a question, it tends to go into a loop of reacting to its own beep unless you have listen_while_talking turned off or good echo cancellation.
Expected Behavior
If Naomi hears audio while active listening, but that audio contains no words, Naomi should not produce the "boop" sound effect or otherwise react to the audio. It should wait until it gets a transcription with words.
Actual Behavior
Naomi beeps to tell the user that it is ready to hear their response to the question just asked, and that beep gets picked up as audio by the vad. Naomi then boops to let the user know that it has captured some audio and is processing it. The audio comes back blank, so Naomi beeps again to let the user know it is ready to listen, but once again picks up the beep and plays a boop. It just keeps beeping and booping until the user finally says something it can work with.
Possible Fix
A better VAD could help, or cutting off the microphone while playing beeps and boops. The best would probably be to hold off on the boop signal until the transcription is returned and checked for content.
Steps to Reproduce
Ask Naomi to tell you about today's news
Naomi will read the top headlines, then offer to email links to the articles to you, then plays a beep and waits for your response.
Before you can respond, Naomi plays a boop noise and starts processing the beep
Description
When Naomi asks a question, it tends to go into a loop of reacting to its own beep unless you have listen_while_talking turned off or good echo cancellation.
Expected Behavior
If Naomi hears audio while active listening, but that audio contains no words, Naomi should not produce the "boop" sound effect or otherwise react to the audio. It should wait until it gets a transcription with words.
Actual Behavior
Naomi beeps to tell the user that it is ready to hear their response to the question just asked, and that beep gets picked up as audio by the vad. Naomi then boops to let the user know that it has captured some audio and is processing it. The audio comes back blank, so Naomi beeps again to let the user know it is ready to listen, but once again picks up the beep and plays a boop. It just keeps beeping and booping until the user finally says something it can work with.
Possible Fix
A better VAD could help, or cutting off the microphone while playing beeps and boops. The best would probably be to hold off on the boop signal until the transcription is returned and checked for content.
Steps to Reproduce