Hey saw your comment on steam. I'll paste my response here as well:
Thanks for the details. Which voice are you using? For testing, please select “Best All-Round Voice (Supertonic)” in Quick Voice. Voice cloning, especially “Expressive HQ Voice + Cloning (High Specs),” can add delays on your hardware.
On some tested RTX 2070/4050 devices, replies take around 5–20 seconds, so 50+ seconds needs investigating. GPU Mode/CUDA doesn't guarantee full GPU offloading. Vision mode, Enhanced Memory Search and other system load can also slow responses.
Could you email Player.log after a slow reply to gaming@elushis.com? Paste this path into File Explorer:
%USERPROFILE%\AppData\LocalLow\Elushis Music _ Gaming\CielChan Desktop AI\Player.log
Ideally, the log should include startup and a slow reply. If possible, also test a plain-text reply with GPU Mode off and let me know how long CPU-only takes.
Thanks again for your help!