Krea2 Edit (source patch):
Krea2EditModelPatch is a specialized node designed to enhance the Krea2 model by integrating an in-context source-preservation path. This node is particularly beneficial for AI artists who wish to maintain the original appearance of a source image while applying edits. It achieves this by utilizing the VAE-encoded source latent as a frame of clean tokens, which are then distinguished from the target tokens using a unique 3-axis RoPE frame index. This method allows for precise control over the editing process, ensuring that the source image's characteristics are preserved while the target image is generated. By wrapping the model's DIFFUSION_MODEL forward, Krea2EditModelPatch reconstructs the sequence to include both the source and target frames, mirroring the ai-toolkit's predict_velocity_edit functionality. This approach provides a seamless integration of source and target elements, making it an essential tool for artists looking to create complex, layered compositions.
Krea2 Edit (source patch) Input Parameters:
model
The model parameter refers to the Krea2 model that will be patched. It is crucial as it serves as the base model to which the source-preservation path is added. This parameter does not have specific minimum, maximum, or default values, as it depends on the user's selection of the Krea2 model they wish to enhance.
source_latent
The source_latent parameter is the VAE-encoded representation of the source image. It plays a vital role in preserving the original appearance of the source image during the editing process. This parameter ensures that the source image's characteristics are maintained while the target image is generated. Like the model parameter, it does not have predefined values and is determined by the user's input.
Krea2 Edit (source patch) Output Parameters:
patched_model
The patched_model is the output of the Krea2EditModelPatch node. It represents the enhanced Krea2 model with the integrated source-preservation path. This output is crucial for artists as it allows them to generate images that maintain the original source's appearance while incorporating new elements. The patched_model provides a seamless blend of source and target features, enabling the creation of complex and visually appealing compositions.
Krea2 Edit (source patch) Usage Tips:
- Ensure that the source image is accurately encoded using the VAE to maintain its characteristics during the editing process.
- Experiment with different Krea2 models to find the one that best suits your artistic vision and desired output.
Krea2 Edit (source patch) Common Errors and Solutions:
"The connected model patch is not a Uni3C ControlNet."
- Explanation: This error occurs when the model patch connected is not compatible with the expected Uni3C ControlNet.
- Solution: Verify that the correct model patch is being used and ensure it is compatible with the Krea2EditModelPatch node.
"This Uni3C ControlNet expects a Wan model with dim {}, the loaded model has dim {}."
- Explanation: This error indicates a mismatch between the expected dimensions of the Uni3C ControlNet and the loaded model.
- Solution: Check the dimensions of the loaded model and ensure they match the expected dimensions for the Uni3C ControlNet. Adjust the model or select a compatible one if necessary.