Next generation face swapper and enhancer
facefusion.ini
overrides to prevent crashesnvidia-smi
hwaccel
which surprisingly improves the FFmpeg performance--output-video-resolution
from not being applieddeoldify
models to the frame colorizerconditional_download
while model removalddcolor
and deoldify
modelgpen_bfr_1024
and gpen_bfr_2048
modelreal_esrgan_x2
and real_hatgan_x4
model--force-download
to download every model at onceinswapper_128_fp16
model fix--skip-download
when models are absenth264_nvenc
and hevc_nvenc
presets mappingconcurrency_count
for ideal Gradio performancescrfd
face detector model--face-landmarker-score
argument--output-image-resolution
argument--log-level debug
is enabledh264_amf
and hevc_amf
video encoders--output-path
behaviour and normalization--temp-frame-quality
argument--trim-frame-start
wave2lip
uniface_256
yoloface
as the default face detector modelage
and gender
to the face debugger items--output-video-resolution
after upscalingrestoreformer
with the latest restoreformer_plus_plus
modelfacefusion.ini
file is missing a section--video-memory-strategy tolerant
facefusion.ini
fileinswapper_128_fp16
model as default--video-memory-strategy
to trade VRAM usage for speedbmp
to the --temp-frame-format
argument--output-video-preset
which defaults to veryfast
--output-video-fps argument
--output-video-resolution
argument--max-memory
to --system-memory-limit
directml
under Windows