I want to develop an AI model for the Nuke Compositing Program. Initially, I’m aiming for a question-answer based system, and I’m currently collecting both synthetic and natural datasets.
I’m considering two approaches:
- Developing a model from scratch
- Taking an existing model and fine-tuning it
In later stages, I plan to have the model write code and perform node manipulations.
Which approach would you recommend? I would appreciate if you could share your experiences and suggestions.