How to Train a Multi-Concept Anima LoRA? How Many Images Needed?

Witty_Mycologist_995 · reddit · 2026-08-03

A Reddit user asks how to train a multi-concept LoRA, e.g., with 4 different tags. They wonder if they need 80 images per tag (320 total), whether regularization is needed, how to vary the dataset, and whether to train with tags or natural language.

Original post →

More from Multimodal

Multimodal channel →