• Add VPP filter preview. ( With additionally included HWEncVppPreview )
• Add --seek support for avi/avs/vpy input.
• Speed up --vpp-kfm.
• Speed up the search prefilter of --vpp-rtgmc.
• Speed up --vpp-degrain.
• Speed up y4m/raw output writes.
• Prefer TensorRT for --vpp-onnx provider=auto, falling back to CUDA. ( #791 )
• Improve timestamp output of --vpp-afs and --vpp-kfm mode=24.
• Fix --vpp-afs timestamps drifting with --avsync vfr or a custom --timebase.
• Fix --vpp-colorspace reporting no path when range and matrix (or similar) change in the same stage.
• Fix a hang and growing memory when aborting during pipeline flush.
• Fix a race that reported "Decode thread is not responding" when the decode thread exited normally while waiting.
• Fix forced termination of the GPU counter thread. ( #809 )