Corneo's Arm Grab Doggystyle Embedding - CivArchive (CivitAI Archive)

Corneo's Arm Grab Doggystyle Embedding - 1.0.0

NSFW

This is a textual inversion embedding to generate doggystyle sex positions with a focus on the arm(s) being pulled back, from a side-on/profile view. Use corneo_side_doggy to activate.

This is a lot more temperamental than my other sex position embeddings (hands are really difficult), but when it does work the results can be great. It was trained exclusively with arm-grab sex facing to the left side of the image, so the vast majority will generate that way but you can get some others in there depending on your tags. You can always flip the image if you want it facing the other way. An interesting quirk of the way this embedding works is that the position will depend on the width of the image - tall images will have the position be more upright, while wider ones will be more horizontal/doggystyle. It can also do spitroast mmf threesome with a little encouragement.

Example images produced with OrangeCocoa and featuring characters from my other embeds, you can find both on my profile https://civarchive.com/user/Master_Corneo

If you enjoy this embed, I genuinely love seeing your creations with it - you can attach them to a review if you're not aware (but not comments sadly)!

Description

FAQ

Comments (30)

yue_liangJan 15, 2023· 7 reactions

CivitAI

You are simply the best.

Master_Corneo

Author

Jan 15, 2023

Thanks a lot! :D

dasorJan 15, 2023

CivitAI

Where did you find the myhands hypernetwork? Is that downloadable?

Master_Corneo

Author

Jan 15, 2023

The hypernetwork I'm mostly using - sadly no I was just given a file on Discord to get hold of it, I don't know where it comes from online to provide a link to

JJoker375Jan 15, 2023· 2 reactions

https://drive.google.com/file/d/1RuwSvE11yDI-kMAXiTMQ8oBCGMUECyLX/view?usp=sharing

mykeehuJan 15, 2023· 1 reaction

CivitAI

I really like your models! On average how many images use for source and what LR do you train on to keep the poses?

Master_Corneo

Author

Jan 15, 2023· 1 reaction

Thanks! I aim for more than 150 images but sometimes it's as low as 80 or so. Usually 0.0005 LR first 5k steps then reducing from there depending how long it's trained for, but I haven't really figured out a good point to train until and sometimes overtrain a bit too and roll back to an earlier number of steps backup

mykeehuJan 15, 2023

Thank you for the information! I was thinking of training poses on TI instead of Dreambooth, but so far it hasn't worked, the end result is constantly taking shape. I've tried with fewer images and didn't think it would take so long to train at 0.0005, but I'll rethink the strategy then. It always burned in before it learned the end result.

Master_Corneo

Author

Jan 15, 2023· 1 reaction

It does take a lot longer than dreambooth, and Lora is the new thing that looks really good and better but I've not tried it yet

mykeehuJan 15, 2023· 1 reaction

I'm interested in Lora too, but the Dreambooth extension is still being put together and improved, so it'll have to wait a bit. In DB I've already solved what I wanted, but the goal is to make the pose work under any model and style, and TI seemed better for that. And the Lora remains to be seen...

NikJoshJan 15, 2023· 3 reactions

CivitAI

How do I use this file in stable diffusion?

Master_Corneo

Author

Jan 15, 2023· 3 reactions

If you're using the automatic1111 web UI, you put the file in the embeddings folder and then use the filename (without extension) as part of the prompt

NikJoshJan 16, 2023

Thank you very much!

solemnsilverwolf619Jan 15, 2023

CivitAI

How do I load this with Automatic1111?

Master_Corneo

Author

Jan 15, 2023· 4 reactions

You put it in the embeddings folder, then use the trigger phrase, which is the filename without the file extension, as part of your prompt

Creative_AtaraxiaJan 26, 2023

@Master_Corneo but with which model?

Master_Corneo

Author

Jan 26, 2023

Any model that's aware of danbooru tags, so practically all the anime ones

Creative_AtaraxiaFeb 2, 2023

@Master_Corneo got it, thanks

atmogenicJan 16, 2023· 3 reactions

CivitAI

amazing work, do you think you can do a threesome next where a girl is sucking c*ck and getting banged from behind?

Master_Corneo

Author

Jan 16, 2023

This embed will actually do that with a little prompting! Try spitroast and mmf_threesome with this one

atmogenicJan 16, 2023

oh nice, i normally do gangbang and threesome but 99% of the time it doesn't work at all and that includes adding 1boy,2boys and such

Master_Corneo

Author

Jan 16, 2023

It's more of an accident here as I couldn't find enough 1boy1girl images as training data so there's a few threesomes in the training data so you can bring it out of the embedding on purpose, or get it randomly sometimes

strangePasserFeb 20, 2023· 1 reaction

CivitAI

Any idea that there is only one person in my output? I have added (((1 boy))). I also add (((solo))) in neg. prompt. I used anything 3.0 and also your checkpoint.

kurofuku900Mar 26, 2023

CivitAI

Does it work for stuff involving the forbidden L word? Because I've been trying and so far nothing at all. Gives some weird stuff

T0pC4tApr 2, 2023· 11 reactions

CivitAI

Currently does not work at all.

icingqueenMay 15, 2023· 1 reaction

CivitAI

could this be made into a lora?

Master_Corneo

Author

Mar 6, 2024

I've added the training data now if anyone wants to do that.

FoxMccloud2022Mar 27, 2024· 1 reaction

CivitAI

I love your embedding model. ^.^ Can't wait to use it soon.

Master_Corneo

Author

Mar 27, 2024

Thanks! It's pretty old now too

FoxMccloud2022Mar 27, 2024

@Master_Corneo NP and sure it may be old. But, the data you trained mixed with chkpt a good lora, and prompts could easily go a long way with your embedding file.

TextualInversion

SD 1.5

by Master_Corneo

Download (Beta) View on CivitAI