I built QualiSVT: A GUI video encoder inspired by ab-av1, featuring SSIMULACRA2, Butteraugli, and CVVDP metrics.
Body:
Hey everyone,
I wanted to share an open-source batch video encoder and quality assessment tool I’ve been developing called QualiSVT. It's primarily optimized for SVT-AV1 and x265, using FFmpeg or HandBrakeCLI as the backend.
One of my favorite tools is ab-av1, so I implemented a similar Auto-CRF search algorithm inside this GUI. It takes samples from your video and uses binary search/interpolation to find the highest CRF that meets your target quality score, while letting you set a maximum file size limit.
However, instead of relying only on VMAF or XPSNR for quality estimation, I integrated support for newer perceptual metrics: SSIMULACRA2, Butteraugli, and CVVDP. (This is hardware-accelerated using FFVship so it runs relatively fast on the GPU).
Some other useful features included:
- Pre-encode estimations (predicts file size, time, and quality before doing a full encode).
- Post-encode visual validation (automatically extracts the worst, mid, and best frames as WebP images so you can compare them side-by-side).
- Power and energy tracking (hooks into Open Hardware Monitor to log your CPU's W and Wh usage during the encode).
- A solid auto-resume feature if your PC crashes or loses power.
If anyone is interested in testing it out or looking at the code, you can find it on GitHub here: [Insert your GitHub link here]
Feedback and bug reports are welcome.