• Run multiple face detectors at once
  • Prevent the 68 to 5 landmark transformation on low confidence
  • Add the high performance scrfd face detector model
  • Render tabs when using multiple UI layouts
  • Gracefully start and stop processing
  • Introduce --output-image-resolution argument
  • Fix green line when using a target image with uneven pixels
  • Introduce ONNX powered frame enhancers
  • Add --face-landmarker-score argument
  • Print internal statistics when --log-level debug is enabled
  • Change the --output-path behaviour and normalization


  • Cloud processing for heavy load models
  • Add another lip-syncer model
  • Extract vocals from audio for better lip-syncer results
  • Audio processor to mix the volume of the audio sources
  • Improve model lookup using checksums


  • Introduce resumable jobs to support partial and batch processing
  • Different return codes on headless mode for success and error
  • Detect webcam devices via ffmpeg to make them selectable
  • Add webcam via IP support


  • Rewrite the UI based on another framework like
Last modified 1h ago