Type | |
Stats | 8,935 90,797 |
Reviews | (496) |
Published | Jun 17, 2023 |
Base Model | |
Training | Steps: 5,000 Epochs: 5 |
Trigger Words | sleepeace |
Training Images | Download |
Hash | AutoV2 98816E9787 |
Trigger word
sleepeace
Introduction
Uttering the word 'sleepeace' instantaneously transforms you into a picture of blissful slumber. This is Version 2 of LorA, featuring a delightfully relaxed mouth that is simply adorable. Different weights can be used to achieve a variety of expressions.
0.8 - 1.0: Any child will fall asleep instantly, dreaming sweet dreams.
0.4 - 0.7: There will be times when you won't want to put them to sleep. You'll be able to see a happy expression.
1.3 and above: No prompt can disturb this peaceful sleep.
Improvements in Version 2
We've improved the issues with hand rendering.
Issues with yellow patterns appearing around the eyes and hair have been addressed.
We've resolved problems with breakdowns when generating at resolutions above 512.
Adjustments were made to more faithfully adhere to prompts.
We've increased the amount of teaching material to accommodate more compositions.
We've tackled the issue of signs of overfitting appearing in several models.
ver LECO
In the SleePeace utilizing LECO, the following characteristics are present:
There is no influence of artistic style on the model.
Even when high weights are applied, images do not break down.
When using it, you need to be cautious of the following points:
You can apply a weight ranging from 1 ~ 5.
The effects applied depend on the weight of the prompt that the model holds, so try
adjusting the weight to apply for each model.
Too many tokens will reduce the effect of LoRA. Adjust by increasing the weight to apply.
In conclusion
As our main concept revolves around 'becoming blissfully asleep in bed', in most cases, the composition will depict sleeping in bed. While this is the intended effect, there may be occasions when you'd want a blissful sleeping face outside of a bed. In Version 2, we've adjusted the teacher images to accommodate such needs. We hope this will be of great assistance to you.
Additionally, while we've created Version 2 to consistently output better results, the 'sleepeace' that Version 1 occasionally displays is absolutely splendid, and there may be instances where it outperforms Version 2. Both are our favorite LorAs, so choosing either one depending on the situation could yield excellent results. We hope this will be of great assistance to you.