Negative Embedding for Deliberate - CivArchive (CivitAI Archive)

Negative Embedding for Deliberate - v1.0-16T

A negative prompt embedding for Deliberate V2. Use in your negative prompt to make the image look better. Do not use as a regular prompt.

Trained on the bad image dataset from here:

https://civarchive.com/models/17083?modelVersionId=20387

Description

Initial embedding trained on a merge based on Deliberate V2, uses 16 tokens.

FAQ

Comments (7)

Redigleb_Doppler2482Apr 5, 2023· 3 reactions

CivitAI

Definitely gonna try both for negative prompt and for postive prompt (i'm an absolute madman!)

shadow0Apr 5, 2023

CivitAI

I've been testing it and it seems interesting, but I'd really like to understand better what it is expected to impact the most. People? Places? Objects?

skewbed

Author

Apr 7, 2023

It tends to make images more detailed or realistic. I've also seen it help with missing limbs and similar issues.

shadow0Apr 8, 2023

@skewbed Thanks. That has been my experience so far... a little more consistent across the board. It's subtle, but definitely there.

OtherPerspectiveJul 1, 2023

CivitAI

Would you be willing to share your process for creating this? I am attempting something similar but not having any luck

skewbed

Author

Jul 2, 2023· 1 reaction

I trained a TI on the dataset from here: https://civitai.com/models/17083?modelVersionId=20387

OtherPerspectiveJul 3, 2023

@skewbed did you use all of the images there? What about captions? Did you need to use any specific settings?

I have looked through that dataset a bit and it seems quite a lot more generalized than I'd like, and much bigger. I am also trying to use custom data

TextualInversion

SD 1.5

by skewbed

Download (Beta) View on CivitAI