Next generation face swapper and enhancer
directml
under Windowsdebug
, info
, warn
and error
log levelsonnxruntime-rocm
support to installerblendswap_256
, simswap_256
and simswap_512_unofficial
to the face swappergpen_bfr_256
and restoreformer
to the face enhancerretinaface
or yunet
modelswebp
images and webm
videosbest-worst
and worst-best
to --face-analyser-order
face_debugger
to visualize the bounding box, key points and face maskfacefusion-pinokio
repository to our GitHub organization--face-mask-blur
and --face-mask-padding
arguments--face-detector-size
and --face-detector-score
argumentsone
modeinsightface
dependency due handcrafted frame processingtensorflow
and PIL
dependencies duo handcrafted frame analysingudp
and v4l2
now with solid 25fps at 1080p--face-recognition
to --face-selector-mode
--face-analyser-direction
to --face-analyser-order
--reference-face-distance
calculation to a 0-1 range--output-path
even without specified --source-path
default
and cpu
choices for the installerfp16
version to the face swapperCodeFormer
, GFPGAN
and GPEN
to the face enhancerRealESRGAN_x2plus
and RealESRNet_x4plus
to the frame enhancer--help
and documentation with arguments grouping--torch
argument to the installeronnxruntime
--skip-download
to skip automated downloadsonnxruntime
version--onnxruntime
to install.py
facefusion-colab
repositoryCUDA
and ROCM
Dockerfile to utilize the GPUudp
and v4l2
streaming--headless
mode--output-path
via the UI--output-image-quality
option for image outputcrf
range for libvpx
video encoderdetect_fps
to work with different encoding