The attached video shows the overall aim: ML vector Roto
IN the folder there is a script that gets the JSON data into Nuke using Python.
But I have been unable to make a GMask using anything in Nuke.
If you could give me some pointers that would be awesome.
ChatGPT just gave up
Python in Flame is a sandboxed automation layer :
Node graph
Attributes
Batch wiring
File I/O
But geometry lives entirely outside that sandbox .
The reason your introspection showed only flame.PyNode is because:
This is by design, not an oversight.
4 Likes
Sinan
December 22, 2025, 12:15am
3
I’m guessing he mistakenly pasted the conversation he had with chatGPT about Nuke.
Here’s @samhodge_aiml ‘s message on Discord
This is what the LLM served up but it was tripping hard, it ended up confessing that it was licked: Why Python can’t do it either (now fully explained)
Python in Flame is a sandboxed automation layer:
Node graph
Attributes
Batch wiring
File I/O
But geometry lives entirely outside that sandbox.
The reason your introspection showed only flame.PyNode is because:
Python sees the node shell
Geometry is owned by a private C++ object
No bridge exists
This is by design, not an oversight.
Sorry brain partially disengaged. I was juggling a few thoughts while running out the door to a health appointment.
I have been unable to replicate the results that I have made in Nuke using the following script:
https://drive.google.com/file/d/1TU1RlEd_9Yi-9l3ebT1zpupvRxkO6akX/view?usp=drive_link that works well in Nuke
With this data file
Which is based on a few frames of this Pexels clip
https://www.pexels.com/video/a-man-on-a-skateboard-is-riding-under-a-bridge-4612888/
So basically I am asking what would it need to import this data into Flame as a GMask?
not sure if it helps , but here is a “raw dump” of a gmask tracer node (no keyframes) :
gmastracer.raw.zip (1.1 KB)
