wait_
New Member
- Feb 6, 2019
- 8
- 65
You don't have permission to view the spoiler content.
Log in or register now.
You don't have permission to view the spoiler content.
Log in or register now.
wait_
New Member
- Feb 6, 2019
- 8
- 65
wait_
New Member
- Feb 6, 2019
- 8
- 65
You don't have permission to view the spoiler content.
Log in or register now.
Trained small model for Kira with dataset of 7 images and she looks like should)
This examples generated with ADetailer.
Without Adetaler, strength = 0.9..................................With Adetaler (face only), strength = 0.9.
Model is mediocre at best, but better than nothing.
Model is mediocre at best, but better than nothing.
Posted:
Pony:
SDXL1.0:
Some explanataion for Kira:
I just wanted to check what if use same dataset and setting to train two models. Did they procduce exact images (with same seed) or not?
I also tryed to retrain v4, but result was complete garbage, so i deleted it (4k buzz for nothing
).
Pony:
You must be registered to see the links
SDXL1.0:
You must be registered to see the links
Some explanataion for Kira:
I just wanted to check what if use same dataset and setting to train two models. Did they procduce exact images (with same seed) or not?
I also tryed to retrain v4, but result was complete garbage, so i deleted it (4k buzz for nothing
).
Last edited:
Made in haste, but for the sake of an example I think it will do...
Which result is better - When all the models are together or is it better separately?
I tested this model only separately.
v.3.5 doesn't look good with close-ups (noise, bad/pale colors) ... sometimes.
I didn't do anything this time, if compared to v.4. Just pick 7 (with captions from v4) images and start training with default (exept shuffling=on) settings on civitai.