How to properly caption images for Krea2 Character LoRA training?

Strange-Knowledge460 · reddit · 2026-08-27

A Reddit user is asking for the correct captioning method when training Krea2 Character LoRAs. There are conflicting approaches: one describes elements that shouldn't be consistent (e.g., pose, background) after the trigger word, while the other suggests only describing features that must remain consistent. The author found success with the former method but seeks the community's consensus on best practices.

Original post →

More from Apps

Apps channel →