GenFill

Generative fill: replace the masked region from a text prompt.

raw .mdPress ? for keys

Node	GenFill
Menu	AI › GenFill
Names in each format
.oink comp (JSON)	`"type": "OC_N_genfill"`
.nk script	none: written as an Oink node
Python	`nuke.createNode("OC_N_genfill")`
Assistant (MCP)	`create_node(type="GenFill")`
Tab search	`genfill`, or its initials `gf`

GenFill properties

Inputs and outputs

Port	Name	Type
input 0	Image	image
input 1	Mask	image
input 2	Reference	image
output 0	Image	image

Knobs

Panel tab Genfill, in panel order.

Knob	Label	Type	Default	Range / values	.nk knob	Description
`prompt`	prompt	text
`negative`	negative	text
`seed`	seed	float	`1`	0 to 2147483647 (integer)
`strength`	strength	float	`1`	0 to 1		1 = ignore what is under the mask; lower keeps its layout
`steps`	steps (0 = default)	float	`0`	0 to 100 (integer)		0 = model default
`guidance`	guidance (0 = default)	float	`0`	0 to 20		0 = model default
`lightning`	Qwen lightning (8 steps)	bool	`true`			Qwen-Image-Edit: 8-step Lightning LoRA (Apache-2.0)
`mask_channel`	mask channel	menu	`alpha`	alpha, red, green, blue, luminance
`invert_mask`	invert mask	bool	`false`
`grow`	grow (px)	float	`8`	0 to 256		Dilate the mask (px)
`feather`	feather (px)	float	`6`	0 to 256		Soft blend edge outside the grown mask (px)
`context`	context	float	`0.75`	0 to 4		Context around the hole the model sees (fraction of its size)
`megapixels`	model megapixels	float	`1`	0.25 to 2.5
`reference_follows`	reference per frame	bool	`true`			Use the Reference input at the same frame (off: a fixed frame)
`reference_frame`	reference frame	float	`1001`	-100000 to 100000 (integer)
`gamut`	working space	menu	`rec709`	rec709 (linear Rec.709), acescg (ACEScg)
`shaper`	encoding	menu	`reinhard`	reinhard, log, srgb
`exposure`	exposure	float	`0`	-6 to 6		Stops before encoding (fixed per shot), removed after
`peak`	peak	float	`16`	1 to 1000		Top of the encoded range: generated pixels can reach it
`regrain`	regrain (match plate grain)	bool	`true`			Add the plate's measured grain to the filled pixels (match shot grain)
`keep_loaded`	keep model loaded	bool	`true`			Keep the model on the GPU after a run so the next one starts warm; the AI worker still unloads it after 5 idle minutes or when another GPU job needs the card

Other knobs (Python, .nk and the assistant; not on the panel)

Knob	Label	Type	Default	Range / values	.nk knob	Description
`gl_color`	Viewer Controls Color	float ×3	`0, 0, 0`	0 to 1
`postage_stamp_frame`	Static Frame	int	`1`	-2147483648 to 2147483647
`lifetimeStart`	Lifetime Start	int	`0`	-2147483648 to 2147483647
`lifetimeEnd`	Lifetime End	int	`100`	-2147483648 to 2147483647
`useLifetime`	Use Lifetime	boolean	`false`
`cached`	Cached	boolean	`false`
`bookmark`	Bookmark	boolean	`false`
`dope_sheet`	Dope Sheet	boolean	`false`
`clone_group`	Clone Group	string
`link_source`	Link Source	string
`link_overrides`	Link Overrides	string
`split_views`	Split Views	string
`model`	Model	string	`auto`			Model id (picked in the panel; licence tier shown there); auto = the fast model that fits the free GPU memory
`results_data`	Results	string	`{}`
`last_data`	Last Run	string	`{}`

AIVectorGenerator

Generate runs optical flow (SEA-RAFT, BSD-3) in the AI worker and caches vectors per frame.

GenVideo

Generative video: text / image / first+last / extension / video-to-video.

On this page

Inputs and outputs Knobs Other knobs (Python, .nk and the assistant; not on the panel)