EasyUse/Inpaintgenerated

Easy Apply PowerPaint(applyPowerPaint)

Applies PowerPaint inpainting to an image using the supplied mask, pipe context, and selected PowerPaint model.

apply Power Paint

pipe
image
mask
pipe
powerpaint_model
COMBO
powerpaint_clip
COMBO
dtype
COMBO
fitting
1
function
COMBO
scale
1
start_at
0
end_at
10000
save_memory
COMBO
Easy Use

Description

Applies PowerPaint inpainting to an image using the supplied mask, pipe context, and selected PowerPaint model. It supports several inpainting functions, making it possible to remove objects, fill regions from context, follow text prompts, respect shape hints, or extend the image.

Inputs

NameTypeDescription
pipePIPE_LINEThe Easy-Use pipeline containing model, VAE, conditioning, and other current generation settings.
imageIMAGEThe image to process.
maskMASKThe mask that defines the region to inpaint. White pixels normally mark the area to repaint; black pixels are preserved.
powerpaint_modelCOMBOSelects the PowerPaint model checkpoint to use.
powerpaint_clipCOMBOSelects the CLIP model that matches the selected PowerPaint model.
dtypeCOMBOCompute dtype for the PowerPaint model: float16, bfloat16, float32, or float64.
fittingFLOATHow closely the generated result fits the original content. Default 1, range 0.3 to 1. Lower values give the model more freedom; higher values keep closer to the original image.
functionCOMBOThe PowerPaint inpainting function: text guided, shape guided, object removal, context aware, or image outpainting.
scaleFLOATGuidance scale for the inpainting diffusion process. Default 1, range 0 to 10.
start_atINTStart of the timestep range where PowerPaint is applied, from 0 to 10000. Default 0.
end_atINTEnd of the timestep range where PowerPaint is applied, from 0 to 10000. Default 10000.
save_memoryCOMBOMemory-saving mode: none, auto, or max. Higher settings reduce memory usage during generation at the cost of speed.

Outputs

NameTypeDescription
pipePIPE_LINEThe updated pipeline, ready for further Easy-Use nodes.

Usage Notes

  • Select a powerpaint_model and a matching powerpaint_clip; they must be compatible with each other.
  • The function changes how the mask and prompt are used:
    • text guided uses the prompt from the pipeline to paint the masked region.
    • shape guided follows shape information in the masked area.
    • object removal removes the masked object and fills the area with plausible background.
    • context aware fills the mask from surrounding image context.
    • image outpainting extends the image outward from the masked region.
  • fitting balances adherence to the original image against generative freedom. With 1, the output stays close to the source; lower values allow more variation.
  • start_at and end_at control the denoising timestep interval during which PowerPaint is active. The default 0 to 10000 spans the entire denoising schedule; restricting the interval can be used to apply PowerPaint only to early or late steps.
  • For large images or limited VRAM, set dtype to float16 and save_memory to auto or max.

Comments

Sign in with GitHub to join the discussion.

Loading comments…