QualityScaler update 1.5

New

• New Fused AI model, which combine the two AI models to produce even better upscale (and reduce Tiles/Merging artifacts)
• New behavior of Tiles/Merging function: if the image enters entirely into VRAM, Tiles/Merging function will not be used (increasing final quality of the image)
• Optimized CPU backend: now will use more cpu cores according to cpu specs
• Support for .qt video files

Improvements / bugfix

• Completely rewritten upscale process (now much more stable and error-free)
• Updated core libraries (better performance, better GPU support, more stability, better GPU usage)
  • Pytorch 1.8.2 ->1.11.0
  • Cuda 11.1 -> 11.3
• Other improvements and deep code cleaning

GUI

• New unified total dark palette
• Other GUI improvement