roop-unleashed

Features

• Platform-independant Browser GUI
• Selection of multiple input/output faces in one go
• Many different swapping modes, first detected, face selections, by gender
• Batch processing of images/videos
• Masking of face occluders using text prompts or automatically
• Optional Face Upscaler/Restoration using different enhancers
• Preview swapping from different video frames
• Live Fake Cam using your webcam
• Extras Tab for cutting videos etc.
• Settings - storing configuration for next session
• Theme Support

Disclaimer

Installation

macOS Installation

Usage

• Windows: run the windows_run.bat from the Installer.
• Linux: python run.py
• macOS: sh runMacOS.sh
• Dockerfile:
  shell

  1docker build -t roop-unleashed . && docker run -t \ 2 -p 7860:7860 \ 3 -v ./config.yaml:/app/config.yaml \ 4 -v ./models:/app/models \ 5 -v ./temp:/app/temp \ 6 -v ./output:/app/output \ 7 roop-unleashed

Note: When you run this program for the first time, it will download some models roughly ~2Gb in size.

Changelog

• Added random face selection mode
• Added alternative face swapping model with 128px & 256 px output (ReSwapper)
• Video repair added to Extras Tab
• Updated most packages to newer versions. CUDA >= 12.4 now required!
• Several minor bugfixes and QoL Changes

• Bugfix: Several possible memory leaks
• Added different output modes, e.g. to virtual cam stream
• New swapping mode "All input faces"
• Average total fps displayed and setting for autorun

• Bugfix: Starting roop-unleashed without NVIDIA gpu but cuda option enabled
• Bugfix: Target Faces couldn't be moved left/right
• Bugfix: Enhancement and upscaling working again in virtual cam
• Corrupt videos caught when adding to target files, displaying warning msg
• Source Files Component cleared after face detection to release temp files
• Added masking and mouth restore options to virtual cam

• Hotfix for gradio pydantic issue with fastapi
• Upgraded to Gradio 4.43 hoping it will fix remaining issues
• Added new action when no face detected -> use last swapped
• Specified image format for image controls - opening new tabs on preview images possible again!
• Hardcoded image output format for livecam to jpeg - might be faster than previous webp
• Chain events to be only executed if previous was a success

• Added ability to move input & target faces order
• New CLI Arguments override settings
• Small UI changes to faceswapping tab
• Added mask option and code for restoration of original mouth area
• Updated gradio to v4.42.0
• Added CLI Arguments --server_share and --cuda_device_id
• Added webp image support

• Bugfix: Post-processing after swapping

• Added subsample upscaling to increase swap resolution
• Upgraded gradio

• Bugfix: Unnecessary init every frame in live-cam
• Bugfix: Installer downloading insightface package each run
• Added xseg masking to live-cam
• Added realesrganx2 to frame processors
• Upgraded some requirements
• Added subtypes and different model support to frame processors
• Allow frame processors to change resolutions of videos
• Different OpenCV Cap for MacOS Virtual Cam
• Added complete frame processing to extras tab
• Colorize, upscale and misc filters added

• Bugfix: Face detection bounding box corrupt values at weird angles
• Rewrote mask previewing to work with every model
• Switching mask engines toggles text interactivity
• Clearing target files, resets face selection dropdown
• Massive rewrite of swapping architecture, needed for xseg implementation
• Added DFL Xseg Support for partial face occlusion
• Face masking only runs when there is a face detected
• Removed unnecessary toggle checkbox for text masking

• Bugfix: Installer pulling latest update on first installation
• Bugfix: Regression issue, blurring/erosion missing from face swap
• Exposed erosion and blur amounts to UI
• Using same values for manual masking too

• Bugfix: Workaround for Gradio Slider Change Bug
• Bugfix: CSS Styling to fix Gradio Image Height Bug
• Made face swapping mask offsets resolution independant
• Show offset mask as overlay
• Changed layout for masking

• Updated to Gradio 4.21.0 - requiring many changes under the hood
• New manual masking (draw the mask yourself)
• Extras Tab, streamlined cutting/joining videos
• Re-added face selection by gender (on-demand loading, default turned off)
• Removed unnecessary activate live-cam option
• Added time info to preview frame and changed frame slider event to allow faster changes

• Bugfix: Installer Path Env
• Bugfix: file attributes
• Video processing checks for presence of ffmpeg and displays warning if not found
• Removed gender + age detection to speed up processing. Option removed from UI
• Replaced restoreformer with restoreformer++
• Live Cam recoded to run separate from virtual cam and without blocking controls
• Swapping with only 1 target face allows selecting from several input faces

• Bugfix: wrong access options when creating folders
• New auto rotation of horizontal faces, fixing bad landmark positions (expanded on

  

  )
• Simple VR Option for stereo Images/Movies, best used in selected face mode
• Added RestoreFormer Enhancer - https://github.com/wzhouxiff/RestoreFormer
• Bumped up package versions for onnx/Torch etc.

• Re-added unified padding to face enhancers
• Fixed DMDNet for all resolutions
• Selecting target face now automatically switches swapping mode to selected
• GPU providers are correctly set using the GUI (needs restart currently)
• Local output folder can be opened from page
• Unfinished extras functions disabled for now
• Installer checks out specific commit, allowing to go back to first install
• Updated readme for new gradio version
• Updated Colab

Acknowledgements