In v4 for kchisara I have enriched the dataset via a simple python pipeline. Previous versions can be considered OKish, nevertheless v4 does improve the face similarity with the original character.

Triggerword: kchisara