Flash AV Software Corp.

Stable Diffusion For Delphi and C++Builder is a component that capable utilize the Stable Diffusion cpp library(https://github.com/leejet/stable-diffusion.cpp) , Segment Anything 2 (https://github.com/facebookresearch/sam2) and Biglama (github.com/advimman/lama) to remove anything to generate high quality image with AI modules.

• New:Voice extractor component can separate the voice and background music based on Kim_Vocal_2.onnx
• Face Detector based on yoloface8
• Face Enhancer based on codeformer
• Face Swapper based on insightface
• Replace background component based on BriaAI
• live portrait component based on https://github.com/KwaiVGI/LivePortrait to transfer face motion of video
  • the demo need to install our ffmpeg converter component (use data/lp_src.png , lp_ref.mp4 for test)
• Upgrade to latest stable-diffusion.cpp , support inpaint (check the new ReplaceAnything demo)
• Remove Anything component based on Big lama.
• Added Single Object track component OSTrack component.
  • Note:the demo for Object Track need to install our ffmpeg player component to track the object in video.
• Added segment anything 2 component.
• Support text 2 image
• Support image 2 image
• Support Control Net
• Support PhotoMaker
• Support upscale the image
• GPU accelerated generation.
• Note:Although we ship the cpu only dll. We would recommend the cuda version dll that is about 20 times faster with Nvidia RTX GPU. To test the examples , you have to download the AI modules separately because those files are very huge.