Oink
AIMaterialscam AIMaterials

AIMaterials

Analyze runs Cosmos DiffusionRenderer 7B (NVIDIA Open Model License, optional: accept the licence first, ~29 GB download) over the frames in overlapping 57-frame windows and outputs layers albedo (lin

raw .md

Analyze runs Cosmos DiffusionRenderer 7B (NVIDIA Open Model License, optional: accept the licence first, ~29 GB download) over the frames in overlapping 57-frame windows and outputs layers albedo (linear base colour), N (camera-space normals like AINormals), roughness, metallic and depth (relative). The model runs at 1280x704; the result is upsampled to the plate guided by the plate (spatial / range sigma). A frame takes about 18 s on the test GPU. Tick which layers to make (out_*); keep loaded keeps the model ready for the next Analyze. Feed it to Relight (materials / generative).

NodeAIMaterials
MenuAI › AIMaterials
Names in each format
.oink comp (JSON)"type": "OC_N_aimaterials"
.nk scriptnone: written as an Oink node
Pythonnuke.createNode("OC_N_aimaterials")
Assistant (MCP)create_node(type="AIMaterials")
Tab searchaimaterials, or its initials aim

AIMaterials properties

Inputs and outputs

PortNameType
input 0Imageimage
input 1camscene
output 0Imageimage

Knobs

Panel tab Aimaterials, in panel order.

KnobLabelTypeDefaultRange / values.nk knobDescription
out_albedoalbedobooltrue
out_normalsnormals (N)booltrue
out_roughnessroughnessbooltrue
out_metallicmetallicbooltrue
out_depthdepth (relative)booltrue
encodeinputmenusrgbsrgb (scene linear), gamma (linear (gamma 2.2)), none (display)
stepsstepsfloat151 to 50 (integer)Diffusion steps per pass (DPM-Solver++ 2M; 8 is closer to a 40-step solution than the reference's 15 Euler steps)
chunkwindowmenu5757 (57 frames), 25 (25 frames), 1 (per frame)
overlapoverlapfloat90 to 24 (integer)Frames shared by consecutive windows (matched + cross-faded)
seedseedfloat10000 to 100000 (integer)
precisionprecisionmenuautoauto, bf16 (bf16 (~19 GB)), fp8 (fp8 (~12 GB))
spatial_sigmaupsample spatialfloat10.1 to 4
range_sigmaupsample edgefloat0.080.005 to 2
keep_other_frameskeep other framesboolfalse
keep_loadedkeep model loadedbooltrueKeep 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)

KnobLabelTypeDefaultRange / values.nk knobDescription
gl_colorViewer Controls Colorfloat ×30, 0, 00 to 1
postage_stamp_frameStatic Frameint1-2147483648 to 2147483647
lifetimeStartLifetime Startint0-2147483648 to 2147483647
lifetimeEndLifetime Endint100-2147483648 to 2147483647
useLifetimeUse Lifetimebooleanfalse
cachedCachedbooleanfalse
bookmarkBookmarkbooleanfalse
dope_sheetDope Sheetbooleanfalse
clone_groupClone Groupstring
link_sourceLink Sourcestring
link_overridesLink Overridesstring
split_viewsSplit Viewsstring
results_dataResultsstring{}
analysed_firstAnalysed Firstint0-2147483648 to 2147483647
analysed_lastAnalysed Lastint-1-2147483648 to 2147483647
seconds_per_frameSeconds / Framefloat0-3.4028234663852886e+38 to 3.4028234663852886e+38
peak_vramPeak VRAM (GB)float0-3.4028234663852886e+38 to 3.4028234663852886e+38

On this page