Question about stylization and Comic IP

Great work, thank you so much for your contribution to the open-source community.

I have the following questions:
1. For training on a specific style (e.g., anime style), approximately how many images are needed? Is this primarily done during the pre-training phase, the SFT phase, or the RL phase? (My understanding is that both pre-training and SFT use stylized data, but the images generated by stylization aren't aesthetically pleasing enough. RL is applied to adjust for different stylizations. Is RL necessary?)

2. How many images are needed to memorize an anime IP (e.g., Luffy)? (Is it done using low-resolution pre-training + SFT training to remember the character?)

The issue names no files, tests, or entry points. It asks for guidance on image counts and the roles of pre-training, SFT, and RL for style learning and character memorization, so there is no defined implementation or completion check for a newcomer.

Written by the indexing model from the issue text.