I've read a few threads about ghosting. I am also experiencing random ghosting when my cameras are encoding H.264H, and stuttering with H.265. This happens only during playback. With either encoding, the video looks perfect live in BI, in the camera's web page, and in VLC. It's only when being played back that there are issues. When cameras are configured for H.264, during playback there's random ghosting others have posted about. I can't really find a pattern to it except that if the system working without ghosting before I go to sleep, by the time morning comes around I am reliably seeing ghosting on all three cameras (see below for list of cameras). When cameras are configured for H.265, during playback there's a momentary stutter seemingly exactly once per second. Video is otherwise OK. Unlike the ghosting with H.264, this symptom is pretty consistent. I can see perfect live video and then when I go to immediately play it back, there's stuttering. Cameras are 2x Amcrest 2496, 1x Amcrest 1051. See below for screen shot of 2496's configuration. Different combos of CBR vs. VBR, bit rates, quality, etc. don't affect the symptoms I am seeing. Blue Iris version 188.8.131.52 x64, running as a service. Direct-to-disk recording and "limit decoding unless required" enabled. Intel hardware accelerated decode including "Also BVR" are enabled -- when I am trying H.264. When I switch the cameras to H265, I turn hardware acceleration off. Computer is i7-6700 w/ 16GB RAM and Win10 1903. The machine is dedicated to BI. CPU usage is around 1-2% with cameras set to H.264, and 8% when set to H.265. Having the video be perfect when viewed live but glitchy when played back is confusing -- isn't "direct to disk" supposed to mean that the exact stream being decoded and played live is what gets written into the .bvr file? I've also seen talk about turning h/w acceleration off (for H.264) helping. But if the video was recorded direct-to-disk shouldn't I just be able to turn off "Also BVR" to remove HA from playback? The stream written to disk should be identical with and without hardware acceleration. Suggestions welcome.