SVFI 3.15 Updates

New Features

• Pro: New VFI Model for Anime: IFUNet，by SVFI Team, higher quality with slower speed

• Pro: New De-duplication mode for Anime: “Frame Re-distribution” mode，which is complimentary to “1-N” dedup modes, improving vfi performance of anime

• Pro: New mode for NCNN Super Resolution Models: TTA, which improves grain details of output

• New VFI Model for 3D Footage: IFRNet，A SOTA(for now hhh) vfi model.

• New VFI Model: NCNN DAIN. Herald of modern VFI. Notice that it consumes a lot of VRAM. If you encounters vkQueueSubmit Failed error, please refers to (https://substance3d.adobe.com/documentation/spdoc/gpu-drivers-crash-with-long-computations-128745489.html) for solution

• Dolby Vision mkv support

• "Keep Subtitles" at the homepage

Optimization

• Fix Wrong transfer resolution when using 3x Super Resolution models

• Fix missing subtitles in some cases

• Fix low bitrates for NVENC encoders under crf mode

• Fix crash of some cases

• Improve 1-N Dedup mode on smoothness

• Optimize selection for VFI, SR models and Graphic Cards selection

• Optimize default settings

• Optimize dynamic scale for VFI

• Optimize RAM, VRAM usage

• Fix other known issues