High latency in live interview mode – AI responds too slowly to completed questions

During live interview sessions, Parakeet AI takes too long to recognize that the interviewer has finished speaking.

Issue:

• Even after the interviewer has completely finished asking a question, the transcript continues updating for several seconds instead of generating an answer immediately.

• The AI appears to still be processing the speech while the interviewer is already waiting for my response.

• This delay makes it difficult to use Parakeet AI in real-time interviews because valuable response time is lost.

Expected behavior:

• Detect the end of the interviewer's speech much faster.

• Begin generating an answer immediately after the interviewer stops talking.

• Reduce speech-to-text and answer generation latency for live conversations.

Actual behavior:

• Transcript remains in an "updating" state even after the question has ended.

• There is a noticeable delay before answer generation starts.

• The overall experience feels too slow for real interview scenarios.

Impact:
In technical interviews, even a 2–5 second delay can make it appear that I'm hesitating or don't know the answer. Faster end-of-speech detection and lower response latency would significantly improve the live interview experience.