Easy Apply Inpaint(applyInpaint)

Description

The Easy Apply Inpaint node (applyInpaint) applies mask-based inpainting to an image while carrying an EasyUse PIPE_LINE through the graph. It belongs to the EasyUse/Inpaint category and returns an updated PIPE_LINE, so it is designed to be inserted into an EasyUse pipeline flow rather than used as a standalone output node.

The node takes the input image, a mask, and a pipeline object, then performs an inpainting pass using the selected backend and conditioning strategy. It exposes controls for mask expansion, compute precision, scheduling, and optional noise-mask handling.

Inputs

pipe

• Type: PIPE_LINE (required)
• The EasyUse pipeline object carrying model, VAE, conditioning, latent, and other workflow state. The inpainting operation is applied in the context of this pipeline, and the updated pipeline is returned as the output.

image

• Type: IMAGE (required)
• The image to be inpainted. In most workflows this is the original image containing the object or region to modify.

mask

• Type: MASK (required)
• A single-channel mask indicating the region to inpaint. Nonzero areas are treated as the region to regenerate.

inpaint_mode

• Type: COMBO (required)
• Options: normal, fooocus_inpaint, brushnet_random, brushnet_segmentation, powerpaint

Selects the inpainting backend or strategy:

• normal — standard inpainting behavior.
• fooocus_inpaint — Fooocus-style inpainting.
• brushnet_random — BrushNet inpainting with a random generation pattern.
• brushnet_segmentation — BrushNet inpainting guided by segmentation information.
• powerpaint — PowerPaint-based inpainting, which makes the function selector useful for different task types.

encode

• Type: COMBO (required, default "none")
• Options: none, vae_encode_inpaint, inpaint_model_conditioning, different_diffusion

Controls how the input image and mask are encoded or conditioned for the inpainting pass:

• none — no additional inpainting-specific encoding is applied.
• vae_encode_inpaint — uses a VAE encoding path specialized for inpainting.
• inpaint_model_conditioning — uses conditioning intended for inpaint models.
• different_diffusion — uses a different-diffusion style conditioning/inference strategy.

grow_mask_by

• Type: INT (required, default 6, range 0–64, step 1)
• Number of pixels to expand the mask before inpainting. Growing the mask helps avoid hard edges and gives the model room to blend generated content into the surrounding area.

dtype

• Type: COMBO (required)
• Options: float16, bfloat16, float32, float64
• Precision used for the inpainting computation. float16 is usually fastest and uses the least memory; bfloat16 may be preferred on compatible hardware; float32 and float64 increase numerical precision at a performance cost.

fitting

• Type: FLOAT (required, default 1, range 0.3–1)
• Controls how strongly the inpainted content is fitted to the selected behavior. 1 is the default full fitting value; lower values can be used to reduce the strength of the conditioning or allow a looser result.

function

• Type: COMBO (required)
• Options: text guided, shape guided, object removal, context aware, image outpainting

Selects the semantic behavior of the inpaint operation, particularly for modes that support distinct task types:

• text guided — generation driven by a text prompt.
• shape guided — generation guided by shape information.
• object removal — removing an object and filling the region naturally.
• context aware — context-aware filling of the masked area.
• image outpainting — extending the image beyond its existing borders.

scale

• Type: FLOAT (required, default 1, range 0–10)
• Additional scale multiplier applied to the inpaint operation. 1 is the neutral default; higher values amplify the scale used by the inpainting model.

start_at / end_at

• Type: INT (required; start_at default 0, end_at default 10000, range 0–10000)
• Define the beginning and end of the sampling/denoising range affected by the inpainting operation. The default 0–10000 covers the full schedule. Restricting this range can be useful for light retouching instead of a full regeneration.

noise_mask

• Type: BOOLEAN (optional, default true)
• When enabled, the node uses a noise mask during sampling so the masked inpaint region is diffused from noise while the unmasked area is preserved. Disable this when the selected encoding/conditioning path already handles the mask or when you need to reduce memory usage.

Outputs

PIPE_LINE

• The updated EasyUse pipeline containing the inpainted result and associated state. Use downstream nodes to decode, preview, or save the result.

Usage Notes

• All inputs except noise_mask are required, so the node can perform an inpainting pass directly.
• The node returns only a PIPE_LINE; it does not output a standalone image or mask.
• For PowerPaint-style task selection, set inpaint_mode to powerpaint and choose the desired function.
• Increasing grow_mask_by is a practical way to reduce visible seams when the input mask is tight around the object.
• When using encode modes such as inpaint_model_conditioning or different_diffusion, test noise_mask both on and off because the interaction depends on the inpainting model and sampler.
• start_at and end_at use a 0–10000 scale. A narrow window can be useful for subtle edits: set start_at high enough to skip early structure-forming steps, or set end_at lower to avoid fully overwriting fine details.