# HUGE shoutout to ShootTheSound and Fizgig! I finally nailed Moss char loras — retained record

> Retained reddit record from reddit:StableDiffusion.

Canonical: https://highsignal.app/data/records/ce7dabcd4a3ab136

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:StableDiffusion)  published 2026-10-26 11:21 ingested 2026-10-27 02:38

[original source ↗](https://www.reddit.com/r/StableDiffusion/comments/1xfu2d6/huge_shoutout_to_shootthesound_and_fizgig_i/)

## retained text



```
The title basically says it all... after days of frustrations and horrible results with other training tools, I was finally able to train some Moss characters lora that are actually very, VERY good... they're even better than my previous ZIT and Krea ones, which is no small feat!

IDK if it's due to the recent Fizgig masked version, the dataset preparation tools, the big amount of epochs (40/50) or what else, but the resulting generations not only keep a striking resemblance in facial features and expressions, but they also understand body features and imperfections very well, something I always found very hard to achieve in Krea, for example.

The ability to send all the training package to Runpod with a few clicks is also awesome, with a couple of dollars I can train a whole lora without torturing my poor potato GPU for days.

I'll very gladly share my settings if you like, but they're mostly the default Moss ones in Fizgig.

Thanks to creator, really!

EDIT: you can find Fizgig here: [https://github.com/shootthesound/Fizgig](https://github.com/shootthesound/Fizgig)  
Also, YT channel is full of useful, updated guides: [https://www.youtube.com/@ShootTheSound](https://www.youtube.com/@ShootTheSound)
```



document fetched 2026-10-27 00:32 · canonical https://reddit.com/r/StableDiffusion/comments/1xfu2d6/huge_shoutout_to_shootthesound_and_fizgig_i
