KIRA Studio / Performance reference notes
Reviewed 6 November 2026

MEMORY
KIRA: developer-confirmed single-video benchmark below 300 MB;
reported app-memory reading approximately 250 MB. An earlier multilayer
observation was approximately 448 MB. GPU/shared memory and whole-system
usage are separate metrics. This is not a universal memory cap.

Published competitor reference points:
Final Cut Pro: below 0.70 GB. Premiere Pro: approximately 4 GB.
Larry Jordan's travel-edit observations, April 2023:
https://larryjordan.com/articles/the-duration-of-a-video-project-does-not-affect-ram/

DaVinci Resolve: approximately 4–5.5 GB across two 4K edits on a MacBook Neo:
https://www.tomsguide.com/computing/macbooks/i-tested-the-macbook-neo-for-4k-video-editing-and-it-surprisingly-didnt-suck-until-it-did

Workloads, versions, hardware and memory accounting differ. These are
reference observations, not a matched four-editor test or a savings ratio.

COPIED-FRAME TRAFFIC
KIRA: zero copied video-frame PCIe traffic on the supported playback path.
This metric does not describe total PCIe utilization. A hardware bus counter
was not sampled. Other device traffic, operations and export are outside
this figure. Competitor rates are path dependent and were not measured.
Apple silicon uses shared memory without a CPU/GPU PCIe copy link:
https://developer.apple.com/videos/play/wwdc2020/10686/

SCRUB MEASUREMENTS
Historical native debug test, 29 October 2026; RTX 3050 Laptop GPU;
H.264 720p intra-frame proxy; 30-second phases.
P99.9 successful frame-submission gaps:
Video: approximately 11.2 ms.
Three-template scene: approximately 13.0 ms.
Camera with depth of field: approximately 30.6 ms.
These are submission-cadence measurements, not input-to-display latency
or guaranteed playback FPS. No matched competitor scrub trace is claimed.