lib.convert Module

Converter for Faceswap

class lib.convert.Adjustments(color: ColorAdjust | None = None, mask: MaskAdjust | None = None, seamless: SeamlessAdjust | None = None, sharpening: ScalingAdjust | None = None)

Dataclass to hold the optional processing plugins

Parameters:

  • color (ColorAdjust | None) – The selected color processing plugin. Default: None

  • mask (MaskAdjust | None) – The selected mask processing plugin. Default: None

  • seamless (SeamlessAdjust | None) – The selected mask processing plugin. Default: None

  • sharpening (ScalingAdjust | None) – The selected mask processing plugin. Default: None

class lib.convert.Converter(output_size: int, coverage_ratio: float, centering: CenteringType, draw_transparent: bool, pre_encode: Callable | None, arguments: Namespace, config_file: str | None = None)

The converter is responsible for swapping the original face(s) in a frame with the output of a trained Faceswap model.

Parameters:

  • output_size (int) – The size of the face, in pixels, that is output from the Faceswap model

  • coverage_ratio (float) – The ratio of the training image that was used for training the Faceswap model

  • centering (CenteringType) – The extracted face centering that the model was trained on (“face” or “legacy”)

  • draw_transparent (bool) – Whether the final output should be drawn onto a transparent layer rather than the original frame. Only available with certain writer plugins.

  • pre_encode (Callable | None) – Some writer plugins support the pre-encoding of images prior to saving out. As patching is done in multiple threads, but writing is done in a single thread, it can speed up the process to do any pre-encoding as part of the converter process.

  • arguments (Namespace) – The arguments that were passed to the convert process as generated from Faceswap’s command line arguments

  • config_file (str | None) – Optional location of custom configuration ini file. If None then use the default config location. Default: None

property cli_arguments: Namespace

The command line arguments passed to the convert process

process(in_queue: EventQueue, out_queue: EventQueue)

Main convert process.

Takes items from the in queue, runs the relevant adjustments, patches faces to final frame and outputs patched frame to the out queue.

Parameters:

  • in_queue (EventQueue) – The output from scripts.convert.Predictor. Contains detected faces from the Faceswap model as well as the frame to be patched.

  • out_queue (EventQueue) – The queue to place patched frames into for writing by one of Faceswap’s plugins.convert.writer plugins.

reinitialize() → None

Reinitialize this Converter.

Called as part of the preview tool. Resets all adjustments then loads the plugins as specified in the current config.

Return type:

None

Classes

Adjustments([color, mask, seamless, sharpening])

Dataclass to hold the optional processing plugins

Converter(output_size, coverage_ratio, ...)

The converter is responsible for swapping the original face(s) in a frame with the output of a trained Faceswap model.