Updates for Rope-Opal:

• This next version focuses on the UI. It's completely overhauled and finally looks more like a modern app. Lots of usability improvements and room to add new features.
• Can now set strength to zero. This effectively turns off the swapper in the render pipeline so you can apply the rest of the options to the original image (e.g., upscale the original face w/o swapping)
• Recording library can be set to FFMPEG or OPENCV
• Real-time audio is now available while previewing. Performance that renders slower than the frame rate will cause audio lag
• The Differencing function has been reworked into the pipeline to produce better results. Consequently it currently does not show up in the mask preview.
• Wrestled back some VRAM from the Ruby upgrades
• Faster loading of some models. Upcoming releases will do further optimizations
• Adjusted the use of filtering and antialiasing
• Yolov8 added as a face detection model selection. FF is having good results with it, so looking forward to hearing your thoughts on its behavior in Rope
• Scrollbars!
• Save/load parameters, and reset to defaults. Rope will auto-load your saved parameters when launched.
• Restorers (GFPGAN, etc) now have option to choose the detection alignment method. You can trade speed vs fidelity vs texture. This includes the original Rope method that, although flawed, maintain the face textures.
• Detection score. Adjust how aggressive Rope is at finding faces.
• Added detailed help text in the lower left when hovering over UI elements.
• Added reverse, forward and beginning to timeline control.