Skip to content

Latest commit

 

History

History
62 lines (31 loc) · 2.44 KB

README.md

File metadata and controls

62 lines (31 loc) · 2.44 KB

MLOPS Stage

Note This is a community project not directly affiliated with MLOPs if some nodes become mature enough they will be proposed upstream but all this is mostly experimental!

Only for MLOPs 2.0+

Sets of HDA Extension for the MLOPS toolset from @Bismuth-Consultancy-BV that I made while learning it. Moslty experimental and POC level for now. Other community members are starting to share tools too on the Bismuth Discord Server

Feel free to PR or propose suggestions

Installation

I recommend the package install method, ask if you have issues.

  • Clone this repo somewhere and note the path (--recursive will be probably be needed soon)
  • Edit the sample package file replacing the path value with yours.
  • Restart Houdini.

Nodes

Scene to Controlnet

The heart or SD from Scene, this creates a strip of images from the given camera and objects, feeding them to ControlNet 1.1 (Depth or Normal supported for now). image

SD From Scene

This high level node wraps the whole Camera to Points -> Control Net -> SD Inference -> Upscaled -> Texture Projection & Baking pipeline image

seam_faces.mp4

DeepBump (soon™)

A simple deepbump integration for MLOPs, I still need to properly check all licenses involved, works well in combo with SD from Scene's generated texture maps.

Camera To Points Simple

A small rework of the old camera to points by #leuns3051 (450384314769735682). Mostly useful for embedding it in HDAs as older versions were removed, the current version in MLOPs is using editable parm making it hard to do that. It misses the OGL render pass from the 2.0 version in MLOPs.


random previews from early tests
sd_from_scene.mp4
multiview.mp4
single_node.mp4