Uncategorized – Houdini-and-Chill

[hou-tops] Archived post by fabriciochamon

Here you go, node recommender trained in Houdini.
I’ve reorganized this thing to avoid external python code, so now a topnet scans hip files and stores node connections. In sops I create the ML examples, which are then trained in another topnet, and finally I do model inference in sops again.
added comments so its (hopefully) easy to follow. Worth noting that it builds a list of available node types based on the current houdini session and won’t evaluate HDAs that are not part of current environment (ie: unloaded packages).
First run takes a while since the ml regression top creates a venv with pytorch (5gb!)
regarding inference: It needs a good amount of examples to produce something useful! also its just prototype ofc, would definitely need some work on the UX, etc..

Attachments in this post:
http://fx-td.com/houdiniandchill/wp-content/uploads/discord/20251906/29/25/image.png
http://fx-td.com/houdiniandchill/wp-content/uploads/discord/20251906/29/25/node_recommender.hiplc

[f_ai_lures] Archived post by lwwwwwws

i like filmot.com for doing something similar, it’s a search engine for youtube subtitles mainly but the builtin player shows the subs as a nice little column you can scroll through really fast and click the little button at the start of each line to jump in the video to that point

it’s kind of insane that you can search for like “vellum configure strut” then keep hitting “next clip” and it’ll jump through every single time someone has talked about vellum struts on all of youtube

Attachments in this post:
http://fx-td.com/houdiniandchill/wp-content/uploads/discord/20240612/24/24/Screenshot_2024-12-24_at_13.27.55.png