Answers take too long to appear
Why it happens
The answer model and the transcription engine both affect this, and the batch engines add a fixed wait because they cannot tell when you have finished speaking.
How to fix it
Settings → Audio & Models → AI Intelligence → Answer Model: prefer a fast model. Reasoning models are slower by design.
Settings → Audio & Models → Speech Recognition: choose Cloud Quality, then set Engine to Deepgram Flux (Live). It has the lowest delay, and it uses more credits than Smart Saver.
Shorten your context. Very long résumé and job-description text adds delay to every answer.
Related fixes
- The transcript stops partway through a meeting
- Suggested follow-ups stopped appearing mid-call
- The wrong person is credited with what was said
